SMOS L1 Processor L1c Data Processing Model

Transcription

SMOS L1 Processor L1c Data Processing Model
SMOS L1 Processor L1c Data Processing
Model
Code :
Issue :
Date :
SO-DS-DME-L1PP-0009
2.11
29/11/2012
Name
Function
A. Gutierrez
Project Engineer
R. Castro
Project Engineer
P. Vieira
Project Engineer
Checked by
J. Barbosa
Quality A. Manager
Approved by
J.Barbosa
Project Manager
Prepared by
Signature
DEIMOS Engenharia
Av. D. João II, Lote 1.17, Torre Zen, 10º,
1998-023 Lisboa, PORTUGAL
Tel: +351 21 893 3013
Fax: +351 21 896 9099
E-mail: mailto:[email protected]
© DEIMOS Engenharia 2012
All Rights Reserved. No part of this document may be reproduced, stored in a retrieval system, or transmitted, in any
form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written
permission of DEIMOS Engenharia
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
i
This page intentionally left blank
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
ii
Document Information
Contract Data
Classification
Contract Number: 4000101241/10/I-AM
Contract Issuer: ESA ESRIN
Internal
Public
Industry
Confidential
Internal Distribution
Name
Unit
Copies
Name
Organisation
Copies
Steven Delwart
ESA
1
Jean Claude Debruyn
ESA
1
External Distribution
Archiving
Word Processor:
MS Word 2000
File Name:
SO-DS-DME-L1PP-0009-DPM-L1c.doc
Archive Code:
SO-DS-DME-L1PP-0009
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
iii
Document Status Log
Issue
Change description
Date
0.1
Table of Contents
2005-06-01
1.0
First Version for CDR
2005-06-30
1.1
Update after CDR RIDs
2005-08-31
1.2
Update after L1PP implementation
2006-04-07
1.3
Final update for Phase 2 completion
2006-06-07
2.0
Updates for L1PP V2R
2006-11-17
2.1
Updates for L1PP V3.5
2007-07-15
2.2
Updates for L1PP V4:
2007-11-16
Approved
- Details on browse generation
- Details on TEC interpolation
- Defined baseline for IRI and IGRF models
- Defined Sun Tails flagging strategy
2.3
Updates for L1PP V5R:
2008-03-31
- Corrected details on L1c Browse generation
- Included usage of Mispointing ADF in attitude
initialisation
2.4
Updates for L1PP V6R:
2008-07-31
- Added details on Analytical Footprint computation in
section 3.2.5 and 5.11
- Included extra computation in Attitude initialisation to
derive rotation matrix from BFP to EF
2.5
Final update for Phase 4 completion (L1PP V2.0):
2008-12-12
- Corrected Blackman factor in pseudo-code in section
5.7.4
- Updated Geometric Rotation formulation in sections
3.1.4 and 5.10
- Clarified inputs for fast method of Footprint
computation in section 5.11
-Clarified Gibbs 2 procedure in section 3.2.2 and 5.8.4
2.6
Update for L1PP V2.2:
2009-07-24
- Added Sunglint area flag computation function in section
5.12
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
Date
:
29/11/2012
Issue
:
2.11
Page
:
iv
Issue
Change description
Date
2.7
Update for L1PP V3.4:
2010-05-31
SO-DS-DME-L1PP-0009
Approved
- Updated Faraday rotation angle equation according to
new version of [RD.14]. Sections 3.1.3 and 5.10.4
- Changed integration time applicable in theoretical
radiometric accuracy according to KP3 experimental
results. Sections 3.2.3 and 5.9.4
2.7b
Update for the beginning of the SMOS operations phase:
2010-07-15
- Introduced clarification on theoretical radiometric
accuracy computation for pure and mixed scenes in
section 5.9.4
- Introduced clarification on BFP and MISP rotation angles
computation by using all rotations contained inside each
file in section 5.2.4
2.8
Update for L1PP V3.5:
2010-10-29
- Added RFI Flagging routines
2.9
Update for L1PP V5.0
2011-05-06
- Updated border flag computation with the distance to
the Unit Circle aliases (“suspenders and belt”) in section
5.12
- Changed the sun circle radius flagging value to be
configurable and added explanation on RFI flags in section
5.12
2.10
Update for L1PP V5.5:
2011-11-29
- Added Open Issue regarding apodisation windows
used in L1
2.11
Update for L1PP V6.0
2012-11-29
- Updated RFI Flagging conditions to inherit RFI
snapshot flag from L1b/L1a products instead of
computing it in L1c
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
v
Table of Contents
1. Introduction
1
1.1. Purpose
1
1.2. Scope
1
1.3. Acronyms and Abbreviations
2
1.4. Applicable and Reference Documents
2
1.4.1. Applicable Documents
2
1.4.2. Reference Documents
2
2. L1b to L1c in the SMOS l1 processor
5
2.1. Geolocation module
6
2.2. Ionospheric Correction Module
7
3. Description of Algorithms
3.1. Ionospheric Correction module
8
8
3.1.1. Computation of TEC
8
3.1.2. Computation of Geomagnetic Angles
8
3.1.3. Computation of the Faraday rotation angles
8
3.1.4. Geometrical rotation
9
3.2. Geolocation module
11
3.2.1. Alias Free FOV Computation
12
3.2.2. Pixel Brightness Temperature computation
13
3.2.3. Pixel Radiometric Accuracy computation
13
3.2.4. Pixel Observation Angles computation
14
3.2.5. Pixel Footprint Shape Computation
14
3.3. Apodisation window computation
16
3.3.1. Standard apodisation
16
3.3.2. Strip-Adaptive method
17
3.4. RFI Flags Parameters Computation
20
4. L1b to L1c High Level Description
21
5. Detailed Processing Model
23
5.1. Geolocation module
23
5.1.1. Inputs
25
5.1.2. Outputs
25
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
vi
5.1.3. List of variables
25
5.1.4. Implementation
27
5.1.5. Error Handling
28
5.2. Initialise Attitude
28
5.2.1. Inputs
28
5.2.2. Outputs
28
5.2.3. List of variables
28
5.2.4. Implementation
29
5.2.5. Error Handling
31
5.3. Ionospheric Correction module
32
5.3.1. Inputs
32
5.3.2. Outputs
32
5.3.3. List of variables
32
5.3.4. Implementation
34
5.3.5. Error Handling
36
5.4. Compute EAF-FOV
36
5.4.1. Inputs
36
5.4.2. Outputs
36
5.4.3. List of variables
37
5.4.4. Implementation
37
5.4.5. Error Handling
39
5.5. Retrieve DGG pixel list
39
5.5.1. Inputs
39
5.5.2. Outputs
39
5.5.3. List of variables
39
5.5.4. Implementation
40
5.5.5. Error Handling
41
5.6. Compute Pixel Angles
41
5.6.1. Inputs
41
5.6.2. Outputs
41
5.6.3. List of variables
41
5.6.4. Implementation
42
5.6.5. Error Handling
43
5.7. Apodisation Window Computation
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
43
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
vii
5.7.1. Inputs
43
5.7.2. Outputs
44
5.7.3. List of variables
44
5.7.4. Implementation
45
5.7.5. Error Handling
47
5.8. Compute Brightness Temperature
47
5.8.1. Inputs
47
5.8.2. Outputs
47
5.8.3. List of variables
47
5.8.4. Implementation
48
5.8.5. Error Handling
49
5.9. Compute Radiometric Accuracy
49
5.9.1. Inputs
49
5.9.2. Outputs
50
5.9.3. List of variables
50
5.9.4. Implementation
51
5.9.5. Error Handling
52
5.10. Compute Pixel Rotations
53
5.10.1. Inputs
53
5.10.2. Outputs
53
5.10.3. List of variables
53
5.10.4. Implementation
54
5.10.5. Error Handling
56
5.11. Compute Pixel Footprint
56
5.11.1. Inputs
57
5.11.2. Outputs
57
5.11.3. List of variables
58
5.11.4. Implementation
59
5.11.5. Error Handling
61
5.12. Compute Pixel Flags
61
5.12.1. Inputs
61
5.12.2. Outputs
62
5.12.3. List of variables
62
5.12.4. Implementation
62
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
viii
5.12.5. Error Handling
67
5.13. Product Generation
67
5.13.1. MPH
67
5.13.2. SPH
68
5.13.3. Data Set
69
6. Open Issues
71
7. Annex 1: Global data types, constants And Functions
72
7.1. Constants
72
7.1.1. Instrument constants
72
7.1.2. Mathematical and physical constants
72
7.2. Basic types
73
7.2.1. t_utc_time
71
7.2.2. t_l1a_aocs_data
71
7.2.3. t_l1a_pvt_data
71
7.2.4. t_l1c_baselines
72
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
ix
List of Figures
Figure 1: Design documents .............................................................................................................................. 1
Figure 2: L1 Processor Complete Dataflow ...................................................................................................... 5
Figure 3: Geolocation Module........................................................................................................................... 6
Figure 4: Ionospheric Correction Module ......................................................................................................... 7
Figure 5: Geolocation and projection angles [RD.11] ....................................................................................... 9
Figure 6: Areas in FOV for a 2-D aperture synthesis Y-shaped interferometric radiometer, with an antenna
spacing of 0.875 wavelengths, 32º array tilt, 30º array steering and 755 km platform height [RA.01] .......... 12
Figure 7: Cone defined on the satellite frame and projection on to the Earth. In blue is the original reference
frame on the satellite and in red it is the cone-aligned one. Other quantities are defined in the text. ............. 16
Figure 8: Derived formula for the pixel radius for RFI flagging ..................................................................... 20
Figure 9: Geolocation Module Data Flow ....................................................................................................... 21
Figure 10: Orchestrator Processing Steps ........................................................................................................ 22
Figure 11: Geolocation module processing steps ............................................................................................ 23
Figure 12: Ionospheric Correction module processing steps ........................................................................... 31
Figure 13: Apodisation Window module processing steps ............................................................................. 42
Figure 14: Sun tails areas ................................................................................................................................ 63
Figure 15: u_v baselines ordering ................................................................................................................... 72
List of Tables
Table 1 – Applicable Documents ...................................................................................................................... 2
Table 2 – Reference Documents ........................................................................................................................ 3
Table 3: Reference articles ................................................................................................................................ 4
Table 4: Geolocation module variable list ....................................................................................................... 26
Table 5: Geolocation module error handling ................................................................................................... 27
Table 6: Initialise Attitude variable list ........................................................................................................... 28
Table 7: Ionospheric Correction variable list .................................................................................................. 33
Table 8: Ionospheric Correction module error handling ................................................................................. 35
Table 9: Compute EAF-FOV variable list ....................................................................................................... 36
Table 10: Compute DGG pixels variable list .................................................................................................. 39
Table 11: Compute Pixel Angles variable list ................................................................................................. 41
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
x
Table 12: Apodisation Window computation variable list .............................................................................. 44
Table 13: Apodisation Window computation error handling .......................................................................... 45
Table 14: Compute BT variable list ................................................................................................................ 47
Table 15: Compute Radiometric Accuracy variable list.................................................................................. 49
Table 16: Compute Pixel Rotations variable list ............................................................................................. 53
Table 17: Compute Pixel Footprint variable list.............................................................................................. 57
Table 18: Compute Pixel Flags variable list .................................................................................................... 61
Table 19: Instrument constants ....................................................................................................................... 70
Table 20: Mathematical and physical constants ............................................................................................. 70
Table 21– t_utc_time ....................................................................................................................................... 71
Table 22 - t_l1a_aocs...................................................................................................................................... 71
Table 23 - t_l1a_pvt......................................................................................................................................... 72
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
1 of 73
1. INTRODUCTION
1.1. Purpose
The main purpose of this document is to provide a detailed definition of the processes and algorithms
contained in the L1b to L1c processing module.
The document begins with a general overview of algorithms to be used in the Geolocation, Ionospheric
Correction and Apodisation Window Computation modules, its main tasks and its role in the Level 1
Processor. Following the identification of a high-level architecture, a further top-down decomposition is
presented to cover the whole set of algorithms involved in the L1C data processing which, in turn, will
be described in detail.
1.2. Scope
The scope of this document is to describe the algorithms involved in the processing of the Geolocation,
Ionospheric Correction and Apodisation Window Computation modules, providing both architectural
and functional definitions of the main features identified.
This document is the second design level of the SMOS Level 1 Processor. As shown in the figure
below, it is based on the overall System Concept [RD.8] and, with the other Data Processing Model
documents [AD.7] and [AD.8], describes all the L1 algorithms used.
System Concept
DPM L1a
DPM L1b
DPM L1c
XML API doc
Architecture
Test Tools Design
implementation
Figure 1: Design documents
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
2 of 73
This document should be read together with the System Concept document [RD.8], in order to get a
complete view of the facility.
1.3. Acronyms and Abbreviations
For the list of acronyms, please refer to the “Directory of Acronyms and abbreviations” [AD.4].
1.4. Applicable and Reference Documents
1.4.1. Applicable Documents
Ref.
Code
Title
AD.1
Removed
AD.2
SO-RS-ESA-PLM-0003
AD.3
Removed
AD.4
SO-LI-CASA-PLM-0094
Directory of Acronyms and abbreviations
AD.5
ECSS-E-40B
ECSS E-40 Software Engineering Standards
AD.6
SO-DS-DME-L1PP-0011
SMOS L1 Processor Algorithm Theoretical Baseline
Definition
AD.7
SO-DS-DME-L1PP-0007
SMOS L1 Processor L0 to L1a Data Processing
Model
AD.8
SO-DS-DME-L1PP-0008
SMOS L1 Processor L1a to L1b Data Processing
Model
AD.9
SO-TN-IDR-GS-0005
SMOS Level 1 and Auxiliary Data Products
Specifications
Issue &
Date
3.0
SMOS System Requirements Document
2.10
29/10/10
2.16
29/11/12
2.16
29/11/12
5.24
06/12/12
Table 1 – Applicable Documents
1.4.2. Reference Documents
Ref.
Code/Author
Title
RD.1
EE-MA-DMS-GS-0001
Earth Explorer Mission CFI Software MISSION
CONVENTIONS DOCUMENT
RD.2
PE-TN-ESA-GS-0001
Earth Explorer Ground Segment File Format
Standard
Issue &
Date
3.7.3
07/05/10
1.4
13/06/03
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Ref.
Code/Author
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
3 of 73
Title
Earth Explorer Mission CFI Software GENERAL
SOFTWARE USER MANUAL
Issue &
Date
3.7.3
RD.3
EE-MA-DMS-GS-0002
RD.4
Removed
RD.5
Removed
RD.6
Removed
RD.7
David M. Le Vine & Saji
Abraham
"Faraday rotation and passive microwave remote
sensing of soil moisture from space"
RD.8
SO-DS-DME-L1PP-0006
SMOS L1 System Concept
RD.9
SO-TN-DME-L1PP-0024
SMOS L1 Full Polarisation Data Processing
RD.10
SMOS-DMS-TN-5100
Adaptive Apodisation Function Development
Technical Note
1.2
RD.11
P. Waldteufel, G. Caudal
“About Off-Axis Radiometric Polarimetric
Measurements”, IGARSS
2002
RD.12
INETI-SMOS-1-2004-TN5100-002
“On the use of 2D KAISER Functions on the
SMOS/MIRAS adaptive striping problem”
2004
RD.13
SO-TN-DME-L1PP-0172
Analytical Pixel Footprint Technical Note
RD.14
M. Zundo, B.Duesmann
SO-TN-ESA-GS-5873
On-ground BT Frame of Reference TN
RD.15
SO-TN-DME-L1PP-0241
RFI Algorithms in L1PP
RD.16
SO-TN-IDR-GS-0005
SMOS Level 1 and Auxiliary Data Products
Specifications
07/05/10
2000
2.9
29/10/10
1.6
16/07/07
1.0
18/04/08
3.3
24/05/10
1.2
26/10/11
5.22
29/11/11
Table 2 – Reference Documents
Ref.
RA.01
Article
Sun Effects in 2-D Aperture Synthesis Radiometry Imaging and Their Cancellation (A. Camps,
M. Vall-llossera, N. Duffo, M. Zapata, I. Corbella, F. Torres, and V. Barrena), IEEE
Transactions on Geoscience and Remote Sensing, 42 (6): 1161-1167. ISSN: 0196-2892, 2004
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Ref.
RA.02
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
4 of 73
Article
Direct Least Squares Fitting of Ellipses (Fitzgibbon, A.~W. and Pilu, M. and Fisher, R.~B.),
IEEE Transactions on Pattern Analysis and Machine Intelligence, 21 (5): 476-480. May 1999
Table 3: Reference articles
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
5 of 73
2. L1B TO L1C IN THE SMOS L1 PROCESSOR
The last step of the processing chain shall perform a combination of the Brightness Temperature
snapshot values retrieved previously into swath format. Instead of being ordered by integration time and
containing the info for each pixel on the FOV, they shall be ordered by pixels identified in a Fixed Earth
Grid, each of them containing the Brightness Temperature values related to different incidence angles
and different integration times. This step shall also perform the required apodisation of the BT values.
Figure 2: L1 Processor Complete Dataflow
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
6 of 73
2.1. Geolocation module
The Geolocation module shall compute the BT values and their incidence angles in the direction of the
required Earth fixed points, while at the same time computing the characteristics of the footprint (area
and orientation). This is achieved by applying a Discrete Fourier Transform on the L1b BT frequency
components and particularising over the Fixed Earth grid points. Using “normal” processing, the pixel
size for a same geographical location changes in size in consecutive snapshots, as the apodisation
applied in the DFT is the same regardless of the point position. This effect shall make it necessary to
provide footprint information to the L2 users so that it can be taken into account.
If the optional strip-adaptive module is used, the pixel size remains uniform all along its pass trough the
FOV, as the apodisation coefficients applied within the DFT change with the point position within the
FOV. The DFT allows the apodisation and interpolation to the Earth Fixed grid in the same processing
step.
This module shall compute sets of temperature values, H and V for dual-pol operation (real valued), and
H, V and HV for full-pol operation (HV is complex valued). The BT values are produced in consecutive
integration times. The Ionospheric Correction module will compute the geomagnetic angles necessary
for the rotation correction from instrument to ground and store them as part of the L1c product.
Figure 3: Geolocation Module
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
7 of 73
2.2. Ionospheric Correction Module
Figure 4: Ionospheric Correction Module
This module shall compute one TEC value and one geomagnetic vector value per each snapshot, so that
the subsequent (L2) Faraday correction for each footprint can be computed using these values. It is
considered that TEC and geomagnetic variations are sufficiently small within the scene FOV such that
one mean value of each can be considered as the baseline. The rotation due to the change of reference
from instrument to ground shall also be computed here on a per-pixel basis.
The information required for the correction shall be extracted from specific auxiliary data combined
with HKTM L1a data. It shall be attached internally to the L1c data, in the initial time-ordered list of
snapshots parameters.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
8 of 73
3. DESCRIPTION OF ALGORITHMS
3.1. Ionospheric Correction module
This module shall compute a set of values on a snapshot basis, which are required for computing the
Faraday rotation angles for each pixel within that snapshot. The values are the Total Electron Content
(TEC) and the geomagnetic angles at boresight, which shall be passed to the Geolocation module for
computation of the Faraday rotation angle.
This Faraday rotation angle accounts for the polarisation mixing that the Brightness Temperatures suffer
when passing through the Ionosphere on the path from each pixel to the MIRAS instrument.
The computation of those values shall be performed based on existing models, and as an additional
possibility, the TEC can be obtained from a fast “IGS Combined” product available every 24h.
Ionospheric correction performs then the computation of the Total Electron Content and the
geomagnetic angles. Later on, the values are particularised at each pixel by simply using the incidence
(θg) and azimuth (φg) angles from the spacecraft to the pixel (shown in Fig. 5).
3.1.1. Computation of TEC
As mentioned before, the TEC can be obtained from several sources, the first one is the IRI2001 model
available as well from NSSDC in FORTRAN code, the second one is the IGS combined TEC map
produced by UPC and available daily as Auxiliary Data File at ftp://gage152.upc.es/rapid_iono_igs.
The IRI2001 model is based on the 11-year Sun solar cycle, and requires as input the geodetic longitude
and latitude, the time and the altitude, as well as several configuration parameters. The IGS combined
TEC map is an ASCII file with TEC values over a Mercator grid at an altitude of 450km and measured
every 2 hours. The frequency with which this file is released is every 24 hours. The expected output in
both methods is the TEC value for the given spatiotemporal coordinates expressed in TEC Units
(10+016e-/m2).
3.1.2. Computation of Geomagnetic Angles
The geomagnetic angles are computed using the IGRF 11th generation model, valid until 2015 and
available as an Auxiliary Data File, and the IGRF FORTRAN code available from NSSDC. The
required inputs at any time are the S/C geodetic longitude and latitude, the time in decimal years and the
altitude. The expected outputs shall be the magnetic field strength in Tesla (F), as well as the magnetic
inclination (I) and declination (D) in degrees.
3.1.3. Computation of the Faraday rotation angles
Computation of Faraday rotation (ω ) for each pixel is computed using Eq.84 of [AD.6], but corrected
according to [RD.14] sign convention and reference (i.e. from surface to antenna), with the retrieved
pixel observation angles, TEC and geomagnetic data.
ω = −6950 × F × TEC × sin I + cos I × tan θ g × cos (φg − D ) 
Eq. 1
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
9 of 73
3.1.4. Geometrical rotation
Apart from the Faraday rotation effect on the electromagnetic wave, another effect must be taken into
account due to the change of reference frame between instrument and pixel. The instrument is
measuring the polarised energy along two axes determined by the output ports of the LICEF receivers,
and these axes do not correspond to the Horizontal and Vertical polarisation axes at pixel level. Thus it
is required to compute another angle (geometrical rotation) to relate the reference change per pixel.
3.1.4.1. Waldteufel and Caudal Implementation
The geometrical rotation between the polarisation axes in the antenna frame and the pixel frame are
computed as follows. The diagram below shows the angles necessary to describe a projection of point P
into the antenna frame (SXYZ)
Figure 5: Geolocation and projection angles [RD.11]
Where t is the tilt angle of the instrument (32.5º), X and Y are the polarisation axes in the antenna
frame, and EH and EV are the polarisation axes in the pixel frame. EH is orthogonal to the PS and PO
directions, and EV is orthogonal to EH and PS.
The angles may be computed using the following expressions, as defined in [RD.11]:
θ = arccos sin t sin θ g + cos t cos θ g 
 − sin t cos θ g + cos t sin θ g sin φ g 

sin θ


φ = arcsin 
Eq. 2
Eq. 3
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
10 of 73
 cos t sin θ g − sin t cos θ g sin φ g 

sin θ


ϕ = π − arcsin 
Eq. 4
Where φ is measured from the X axis and positive clockwise in the XYZ reference frame, and ϕ is the in
plane geometrical rotation between polarisation frames. The above expressions are valid for the range
 π π
 π 3π 
φ g ∈  − ,  , for the range φ g ∈  ,  , the angles must be simply computed as:
 2 2
2 2 
 − sin t cos θ g + cos t sin θ g sin φ g 

sin θ


φ = −π − arcsin 
Eq. 5
 cos t sin θ g − sin t cos θ g sin φg 

sin θ


ϕ = arcsin 
Eq. 6
The final geometrical rotation value from pixel frame to antenna frame is given by α = φ − ϕ
The antenna frame observation angles can also be obtained using the Cartesian coordinates of the S/C
and the Cartesian coordinates of the pixel, plus the instrument attitude rotation matrix (and additionally
miss-pointing like Best Fit Plane deviation). Using this last method, the pixel coordinates do not need to
be restricted to the Reference Ellipsoid, but they can be expanded with the local altitude value using a
Data Elevation Model. This process of orthorectification shall be done with the help of the EE CFI
functions.
This method is no longer used since L1PP v2.0, but its description is kept for historical reasons.
3.1.4.2. Duesmann and Zundo Implementation
In [RD.14] an alternative method to compute the geometric rotation angle is proposed. It consists on
calculating the emission basis vectors in the Earth’s surface and the Ludwig-3 polarization basis vectors
on the instrument frame and from them extract the geometrical rotation angle, α. This method is the
algorithm baseline since L1PP v2.0.
The emission basis vectors in the topocentric frame of the pixel are calculated from the observation
angles as
)
Ehextopocentric = − cos φe
)
Eq. 7
Eheytopocentric = sin φe
)
Eheztopocentric = 0
)
Evextopocentric = − cos θ e sin φe
)
Eveytopocentric = − cos θ e cos φe
)
Eveztopocentric = sin θ e
Eq. 8
To calculate the Ludwig-3 basis vectors on the satellite one must start by defining the target vector from
r
the satellite to the pixel, r MIRAS ,
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
 cos θ s sin φs
r MIRAS 
= cos θ s cos φs
r
 − sin θ
s

Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
11 of 73
Eq. 9
and with it define two auxiliary vectors, Nx and Ny:
r
vaux MIRAS = r MIRAS − u z MIRAS
Nx
MIRAS
Ny
MIRAS
vaux MIRAS × u x MIRAS
=
vaux MIRAS × u x MIRAS
=
Eq. 10
vaux MIRAS × u y MIRAS
vaux MIRAS × u y MIRAS
These auxiliary vectors are used to calculate the Ludwig-3 vectors
r
N MIRAS × r MIRAS
Lx MIRAS = x MIRAS r MIRAS
Nx
×r
r
N y MIRAS × r MIRAS
MIRAS
Ly
= MIRAS r MIRAS
Ny
×r
Eq. 11
)
)
Having Ehetopocentric , Evetopocentric , Lx MIRAS and Ly MIRAS to the same referential, the polarization angle is
obtained as shown in the following equation:
r
r
 EhMIRAS ⋅ Ly MIRAS
r MIRAS
α ′ =arctan  r
 Eh
⋅
L
MIRAS
x





Eq. 12
For further details, please refer to [RD.14].
3.2. Geolocation module
The objective of the Geolocation process is to compute for each of the pixels observed the
corresponding Brightness Temperature and observation angles, the radiometric accuracy, the Faraday
and geometric rotation angles, the footprint major and minor semi-axis and a set of quality flags. The
pixels shall be part of an Earth Fixed Grid (ISEA), such that the measurements are always referred to the
same geodetic coordinates.
In order to compute the data, it shall use the L1b Brightness Temperature frequencies to perform a DFT
in the direction of each pixel, plus obtaining a radiometric accuracy measurement of the errors induced
in that measurement. It shall also compute the equivalent synthesized beam in order to extract the pixel
(or footprint) spatial resolution.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
12 of 73
3.2.1. Alias Free FOV Computation
The first step shall be to identify which pixels of the ISEA grid are contained within the extended aliasfree FOV for one particular snapshot. This shall be done by first projecting the extended AF-FOV
contour onto the ISEA grid and gathering the points falling inside.
After identifying which pixels of the ISEA grid are contained within the extended alias-free FOV for
each snapshot, the incidence and azimuth angles in the antenna frame can be computed for each of the
pixels, which also correspond to the xi-eta coordinates, through to the following equation:
ξ = sin θ cos φ
η = sin θ sin φ
Eq. 13
As it will be shown during the implementation, different angle conventions between the usual SMOS
documentation and the CFI angle convention may show different equations. For example in the above
equation θ is measured from the boresight towards the viewing direction, and φ is measured from the 0º
axis.
The computation of the EAF-FOV is done through the selection of a geometrical boundary that follows
certain requirements. These requirements can be observed in the next figure:
Figure 6: Areas in FOV for a 2-D aperture synthesis Y-shaped interferometric radiometer,
with an antenna spacing of 0.875 wavelengths, 32º array tilt, 30º array steering and 755
km platform height [RA.01]
The exact procedure shall be detailed in section 5.
3.2.2. Pixel Brightness Temperature computation
The Brightness Temperature computation to be performed per pixel shall be a Discrete Fourier
Transform, using as input the brightness temperatures frequencies Tˆ u,v produced as L1b output, the
( )
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
13 of 73
(u,v) baselines coordinates, the apodisation window coefficients W(u,v) and the xi-eta coordinates of
that particular point:
T (ξ ,η ) =
3 2
j 2π u ξ + v η
d ∑∑ Tˆ ( umn , vmn ) ⋅ W ( umn , vmn ) ⋅ e ( mn mn )
2
m n
Eq. 14
In the nominal case, the apodisation window W coefficients shall be constant regardless of the xi-eta
coordinates (e.g. Blackman window). In the case of strip adaptive, the coefficient shall be also a
function of the xi-eta coordinates. The method for W computation is shown in the last chapter of this
section.
In case that Gibbs correction was applied during L1b processing, this correction removed a constant flat
Earth or a constant Land and a model Ocean from the reconstructed image, which must be added back
after the DFT is performed. This constant Earth or Land Brightness Temperature is reported inside each
snapshot of the L1b products, and is applicable only for dual polarisation scenes, never for full pol
scenes. The model Ocean needs to be recomputed for each pixel affected depending on the observation
angles, and is the same model that was subtracted in the L1b processing from [AD.8]
3.2.3. Pixel Radiometric Accuracy computation
Additionally, the radiometric accuracy shall be computed for each pixel and polarisation. The equation
for doing this is the following:
∆T
pq
Ω pq 1 − ξ 2 − η 2 3 2
d
(ξ ,η ) =
G pq (ξ ,η )
2
Tsyspq
B ⋅τ eff
α wα ol
Eq. 15
Where the different parameters are:
4π
for the current LICEFs, where
D
D is the averaged directivity of the LICEFs at the corresponding polarisation pq.
Ω pq is the solid angle of the antenna, and can be approximated by
G pq (ξ ,η ) is the averaged LICEF receiver directional power Gain function normalised so that it is
unity at boresight. This data shall be retrieved from the antenna patterns ADF for each polarisation.
d is the distance ratio between receivers (0.875)
Tsyspq is the averaged System Temperature measured by the PMS system, which has been used to de-
normalise the L1a calibrated visibilities.
B is the equivalent receiving frequency bandwidth in Hz, currently being 19 MHz
τ eff is the effective integration time, and is equivalent to τ ceff where τ is the integration time (1.2s
for pure dual pol epochs, 0.46 for mixed dual pol epochs, and 0.64 for HV pol epochs) and ceff is the
coefficient that accounts for the 1-bit correlation, oversampling and hermiticity (ceff=1.81).
αw =
∑∑
u
v
(W ( u, v ) )
R ( u, v )
2
accounts for the apodisation window and redundancies in the
measurements, where W is the apodisation window term for each (u,v) baseline and R is the
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
14 of 73
redundancy level of that same baseline (i.e. number of times that the baseline has been measured, 1
for non-redundant baselines, greater than 1 for the rest)
 f −f 
−2π  0 01 
 B 
2
α ol = 1 + e
accounts for the local oscillator factor, where B is the bandwidth mentioned
before, f0 is the central frequency and f01 is the low frequency, whose values are obtained from the
PLM ADF.
3.2.4. Pixel Observation Angles computation
Afterwards, the incidence (θp) and azimuth (φp) observation angles shall be computed. The incidence
angle is the angle between the local normal at each pixel and the pixel-to-satellite direction, whereas the
azimuth angle is the angle measured between the pixel-to-satellite direction projected in the local
tangent plane and the local North direction. These angles shall be computed with the help of the CFI
functions, using the spacecraft position and the latitude, longitude and altitude coordinates of each pixel
in the Earth fixed grid.
3.2.5. Pixel Footprint Shape Computation
Beyond these computations, the pixel shape shall be obtained as the projection onto the Earth fixed Grid
of the –3dB contour of the Synthetic Antenna Directional Gain.
On a first approximation, this projection shall be approximated by an ellipse, in which its major semiaxis is not necessarily oriented with the azimuth (φp) observation angle. This makes it mandatory to
project at least 6 points in order to make a least squares fit of an ellipse to derive its semi-axes.
The Synthetic Antenna Directional Gain (also known as Equivalent Array Factor) may be computed in
the antenna frame by following the expression:
AFeq (ξ , ξ ′,η ,η ′ ) =
 u ⋅ ξ + vmn ⋅η  j 2π ( umn (ξ −ξ ′)+ vmn (η −η ′) )
3 2
d ∑∑ W ( umn , vmn ) ⋅ r%  − mn
⋅e
2
f
m n

0

Eq. 16
Where:
W ( umn , vmn ) is the apodisation function computed before
r% is the fringe-washing function which accounts for the spatial decorrelation between antennas. It is
 −π W 2 ( uξ + vη )2 
r

calibrated as part of the nominal processing, but it can also be approximated by r% ≈ e 
,
where Wr is the relative bandwidth of the filters (i.e. bandwidth divided by the central frequency)
u ,v
( mn mn ) are the baseline coordinates in the frequency domain
(ξ ′,η ′ ) are the coordinates of the resulting distribution in the antenna frame
d is the antenna element spacing (0.875)
f0 is the central frequency (1413 MHz)
The resulting distribution over the antenna frame, cut at half of the maximum (–3dB) will yield a
contour (circular or elliptical) that must be projected over the Earth. As mentioned, the major semi-axis
may not be oriented along the azimuth (φp) observation angle, so it should be necessary to project at
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
15 of 73
least six points to compute the ellipse axes on the ground by fitting those points to an ellipse quadratic
equation, using a method shown in [RA.02].
As a further simplification, calculating the pixel footprint can be described as the geometrical problem
of finding the intersection of an elliptical cone, defined in a satellite frame, with a plane defined in the
Earth topocentric frame of the target. The low level description of the algorithm can be found in
[RD.13].
Given a looking direction on the satellite, defined by an azimuth (ϕ) and an elevation (θ), a circumference
on the xOy satellite plane (that corresponds to the -3dB contour of the Synthetic Antenna Directional
Gain) and the unit semi-sphere of the satellite an elliptical cone can be defined in a satellite reference
frame. The projection of its base on the target topocentric plane defines an ellipse.
The following figure depicts the geometry of the problem. The meaning of the reference frames
illustrated in red and blue is clarified in Section 5.11.
Figure 7: Cone defined on the satellite frame and projection on to the Earth. In blue is the
original reference frame on the satellite and in red it is the cone-aligned one. Other
quantities are defined in the text.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
16 of 73
3.3. Apodisation window computation
3.3.1. Standard apodisation
In the nominal case, the apodisation window is simply a function of the u,v baseline coordinates.
For the Blackman window, the expression is the following:

W ( u , v ) = 0.42 + 0.5 ⋅ cos  π


u 2 + v2
3 N EL d


 + 0.08 ⋅ cos  2π




u 2 + v2
3 N EL d




Eq. 17
In which the MIRAS/SMOS case for a Y-shaped array has already been taken into account, and NEL=21
and d=0.875.
3.3.2. Strip-Adaptive method
In order to ensure a circular footprint on the ground, the process must be initiated backwards, i.e. given
the Earth pixel; a ground circular footprint is considered and projected onto the antenna frame as an
ellipse. This ellipse shall be the desired output of the –3dB contour for the Array Factor computed in
equation 10, which will require a specific window for that purpose.
As seen in [AD.6] the rotation matrix describing the coordinate transformation between Earth fixed grid
pixel coordinates and antenna plane spherical coordinates is given by:
 dxlP   cos ε cos φ g cos ε sin φ g cos t − sin ε sin t − cos ε sin φ g sin t − sin ε cos t 


 
− cos φ g sin t
cos φ g cos t
×
 dylP  =  − sin φ g
 dz   sin ε cos φ sin ε sin φ cos t + cos ε sin t − sin ε sin φ sin t + cos ε cos t 
g
g
g
 lP  

 r cosθ cos φ −r sin θ sin φ sin θ cos φ   dθ 
 dθ 

  
 
×  r cosθ sin φ r sin θ cos φ sin θ sin φ  ×  dφ  = [ L ] ×  dφ 
 −r sin θ
 dr 
0
cosθ   dr 

 
Eq. 18
In which
(xl,yl,zl) are the pixel coordinates in the Earth fixed grid
ε is the arc between the nadir and the point position from the centre of the Earth assumed as
spherical
φ g is the azimuth angle from the nadir
t is the tilt angle of the antenna
r, θ and φ are the spherical coordinates of the pixel in the antenna frame
L is the complete rotation matrix
Projecting a circular pixel of radius R on the Earth surface into the antenna frame using the rotation
matrix L, results in an ellipse equation expressed in xi-eta coordinates, as demonstrated in [AD.6].
A ⋅∆ξ
2
+ 2 ⋅B ⋅d ∆ξ ⋅∆η + C ⋅∆η
2
= 1
Eq. 19
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
17 of 73
Where the ellipse coefficients are:
A=
1
R2
 2
cos 2 φ
cos φ sin φ
sin 2 φ 
2
2
2
(
L
+
L
)
−
2
⋅
(
L
L
+
L
L
)
+
(
L
+
L
)
 11

21
11 12
21 22
12
22
cos 2 θ
cos θ sin θ
sin 2 θ 

Eq. 20
B=
1 
2 cos 2 φ − 1
cos φ sin φ
cos φ sin φ 
2
2
(
L
L
+
L
L
)
− ( L11
+ L221 )
− ( L12
+ L222 )

11 12
21 22
2 
2
R 
cos θ sin θ
cos θ
sin 2 θ 
Eq. 21
C =
1
R2
 2
sin 2 φ
cos φ sin φ
cos 2 φ 
2
2
2
+
+
⋅
+
+
+
(
L
L
)
2
(
L
L
L
L
)
(
L
L
)
 11

21
11 12
21 22
12
22
cos 2 θ
cos θ sin θ
sin 2 θ 

Eq. 22
The angle of the first semi-axis with respect to the axis ξ is:
 2B 

 A−C 
δ = arctan 
Eq. 23
and the semi-axis are defined by:
E1 = A cos 2 δ + 2 B sin δ cos δ + C sin 2 δ
Eq. 24
A sin 2 δ − 2 B sin δ cos δ + C cos 2 δ
Eq. 25
E2 =
So to circularise each pixel, it is required to compute its elliptical projection in the antenna frame using
Eqs. 17 to 19.
Once we have the desired ellipse parameters in the antenna frame, it is required to find a suitable
apodisation window whose Equivalent Array Factor –3dB contour fits that ellipse.
For this purpose, a previous study described in [RD.10] shows how to adjust a 2D window based in
Kaiser windows. This apodisation window expression is the following:
2
2


 u ′  
 v′  


I0 αu 1 − 
I α 1− 



ρmax   0  v
ρ max  


× 

W ( u, v ) = 
I 0 (α u )
I 0 (α v )
Eq. 26
Where I0 is the Modified Bessel Function of the First Kind, and has the following expression:
2k
( x / 2)
I0 ( x ) = ∑
2
k = 0 ( k !)
∞
Eq. 27
The alpha parameter in each Kaiser window drives the tapering capabilities of the apodisation, meaning
that higher values of alpha provide higher tapering. The baselines are normalised with the same
2
2
maximum value ρ max = umax
+ vmax
= 3N EL d .
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
18 of 73
Rotation is introduced by making a linear combination of the u and v frequencies and applying the
apodisation along those new directions, as it is shown in the following equation:
u ′ = u cos δ + v sin δ
v′ = −u sin δ + v cos δ
Eq. 28
The problem to be solved would be to find the αu and αv coefficients by forcing that the Equivalent
Array factor as computed in Eq. 10 particularised at the semi-axes points needs to be half of the
maximum (measured at xi=eta=0). The expression has been simplified considering FWF unity on the
right side of the equations:
∑∑
u
v
∑∑
u
v
W ( u ′, v′ )
2
W ( u ′, v′ )
2
= ∑∑ W ( u ′, v′ ) × e j 2π uE1
u
v
= ∑∑ W ( u ′, v′ ) × e
u
Eq. 29
j 2π vE2
v
Substituting Eqs. 20 and 22 in the above one, it results in an equation system with two unknowns:
2
2


 u cos δ + v sin δ  
 −u sin δ + v cos δ  
1


  × I0 αv 1 − 
 =
∑∑ I 0 α u 1 − 
2 u v 
ρ max
ρ max



 



2
2


 u cos δ + v sin δ  
 −u sin δ + v cos δ   j 2π uE1


= ∑∑ I 0 α u 1 − 
  × I0 α v 1 − 
 ×e

ρ max
ρ max
u
v



 



2
2


 u cos δ + v sin δ  
 −u sin δ + v cos δ  
1


  × I0 αv 1 − 
 =
∑∑ I 0 α u 1 − 
2 u v 
ρ max
ρ max



 



2 
2 






δ
δ
δ
δ
u
cos
+
v
sin
−
u
sin
+
v
cos
j 2π vE2
= ∑∑ I 0  α u 1 − 
  × I 0  α v 1 − 
  × e

ρ max
ρ max
u
v

 

 


Eq. 30
Which unfortunately cannot be solved analytically, so the solution is to compute the Equivalent Array
Factor according to Eq. 10 using an array of fixed alpha and δ values. Once the Array Factor is
computed, the –3dB contour is fitted with an ellipse, from which the E1, E2 parameters can be obtained.
This procedure described in Eq.24 needs to be done only once, to obtain the parameter distribution of
E1, E2 and δ against αu and αv.
The way to obtain the corresponding apodisation parameters, required for generating an elliptical
footprint in the antenna frame, is then to interpolate in the parameter distribution. Please refer to [AD.6]
to find detailed plots showing the procedure. The numeric implementation shall be described in Section
5.4.
A system of analytical equations has been derived in [RD.12] by INETI after implementation of the
equations above. The system of equations computes the Kaiser parameters (αu and αv) out of the
relationship between the ellipse semi-axes in the antenna frame. The coefficients used in that system of
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
19 of 73
equations are described in the next equation and have been set as part of the Strip Adaptive ADF
(APOD99):
2


α 
w 
 w 
 w 
log10  u  = −7.2205E −8 + 1.9915log10  1  + 1.0776 log10  1   + 0.13022 log10  1  
 αv 
 w2 
 w2  
 w2  


3
2
log10 (α uα v ) = 8.4703 + 1.5081log10 ( w1w2 ) − 0.16293 log10 ( w1w2 )  − 0.016226  log10 ( w1w2 ) 
Eq. 31
3
3.4. RFI Flags Parameters Computation
After the Commissioning Phase, a tailoring of the RFI Flags has been performed. In this sense, there are
three possible mechanisms to flag measurements affected by RFI.
Given a RFI with known coordinates (latitude and longitude) the flagging strategy is defined by the new
AUX_RFI LST ADF. The possible options are:
• To flag the pixels contained in a circle of radius n-pixels around the RFI position;
• To flag the pixels that form the tails (with a width of 2n-pixels) of the RFI (see Figure 14 for
details on the “tail” considered for the direct Sun) ;
• To flag all the pixels from a snapshot as corrupted by RFI (inherited from L1a RFI flags).
The radius of the circle is computed based on empirical data and its formula is presented in Figure 8.
Figure 8: Derived formula for the pixel radius for RFI flagging
The tails are computed as described in Section 5.12.4 using the diameter of the flagged circle as the
threshold to consider a pixel as being affected by the RFI tails.
For more information on the flagging, please refer to [RD.15] and [].
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
20 of 73
4. L1B TO L1C HIGH LEVEL DESCRIPTION
The L1b to L1c processing module shall compute the swath ordered brightness temperatures and the
Ionospheric Correction factors. L1b data comprises the brightness temperature frequency values.
Additional L1a data shall provide the status of the spacecraft (state vector and attitude). Auxiliary data
files provide VTEC, Fixed Earth Grid values for lat, long and altitude, geomagnetic model coefficients,
RFI mask and specific apodisation window coefficients.
The following figure shows the intended data flow.
Figure 9: Geolocation Module Data Flow
L1b to L1c processing module will poll, with a predefined interval for the Level 1b products. Every time
a new product is detected, the reconstructed data will be forwarded to the geolocation module. The
processing is sequential, only the geolocation module needs to be invoked. This module will then
process all available data and return the L1c products for storage.
The footprint info shall be attached to each BT value, along with the incidence angle and special flags.
The data shall be ordered by pixels, where for each pixel the numeric identifier to the ISEA grid and the
number of BT measurements available shall be indicated. Each of the measurements will be composed
of the BT component measured (real for H and V, complex for HV), incidence (θp) and azimuth (φp)
observation angles, radiometric accuracy, Faraday rotation angle (ω), geometric rotation angle (ϕ), pixel
footprint elliptical major and minor semi-axes and quality flags. These flags shall report the position of
the pixel within the FOV (in/out of alias-free FOV, in/out of extended alias-free FOV and close to the
EAF-FOV border or to the unit circle border aliases, also known as “suspenders and belts”) as well as
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
21 of 73
represent any external source that may affect the pixel, like a direct Sun alias, reflected Sun alias, RFI
sources, etc. Input data for the Level 2 processing shall always be a geo-sorted L1c product.
Figure 10: Orchestrator Processing Steps
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
22 of 73
5. DETAILED PROCESSING MODEL
The following sections will detail the implementation of each processing unit, with processing diagrams
and variable lists.
5.1. Geolocation module
The objective of the Geolocation process is first to compute for each scene or integration time the list of
Earth fixed pixels falling inside the Extended Alias-Free FOV. For each of these pixels observed, it shall
then compute the corresponding Brightness Temperature and observation angles, the radiometric
accuracy, the Faraday and geometric rotation angles, the footprint major and minor semi-axis and a set
of quality flags corresponding to that particular time.
The output L1c product pixels will then be formed by aggregations of BT values and their incidence
observation angles for each pixel covered in that swath. Additionally, there shall also be particular
information about the footprint shape and orientation and the snapshot in which the values were
measured.
Using the snapshot reference, there shall be another block of data containing relevant information about
the snapshot, like UTC time, PVT and AOCS, as well as the TEC and geomagnetic model values used
in the Faraday rotation correction.
Different L1c products are generated depending on the polarisation mode of the instrument (dual or full
polarimetric) and with the apodisation window used in the processing (either land or sea).
There shall also be Browse L1c products, containing the same information as nominal L1c, but
providing only one value per pixel considered at an incidence angle of 42.5º.
The processing diagram explains how the process is done, first at the level of the l1b scenes, and later at
the level of each pixel.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
23 of 73
Figure 11: Geolocation module processing steps
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
24 of 73
5.1.1. Inputs
L1b data (SM_xxxx_ MIR_SC_D1B<ID> or SM_xxxx_ MIR_SC_F1B<ID>)
•
ˆ
Reconstructed TB frequencies at top of the atmosphere
•
UTC time
•
PVT and AOCS
L1a HKTM Data (SM_xxxx_TLM_MIRA1A_<ID>)
•
System Temperatures
Auxiliary Data
•
Earth Fixed Grid (SM_xxxx_AUX_DGG___<ID>)
•
Land Sea Mask (SM_xxxx_AUX_MASK__<ID>)
•
RFI mask (SM_xxxx_AUX_RFI___<ID>)
•
Average Antenna Patterns (SM_xxxx_AUX_PATT<ID>)
•
Best Fit Plane (SM_xxxx_AUX_BFP___<ID>)
•
Mispointing Angles (SM_xxxx_AUX_MISP__<ID>)
•
Default PLM values (SM_xxxx_AUX_PLM___<ID>)
5.1.2. Outputs
L1c data (TB swath)
•
Science Data, Land, Dual Polarisation (SM_xxxx_MIR_SCLD1C<ID>)
•
Science Data, Land, Full Polarisation (SM_xxxx_MIR_SCLF1C<ID>)
•
Science Data, Sea, Dual Polarisation (SM_xxxx_MIR_SCSD1C<ID>)
•
Science Data, Sea, Full Polarisation (SM_xxxx_MIR_SCSF1C<ID>)
•
Browse Data, Land, Dual Polarisation (SM_xxxx_MIR_BWLD1C<ID>)
•
Browse Data, Land, Full Polarisation (SM_xxxx_MIR_BWLF1C<ID>)
•
Browse Data, Sea, Dual Polarisation (SM_xxxx_MIR_BWSD1C<ID>)
•
Browse Data, Sea, Full Polarisation (SM_xxxx_MIR_BWSF1C<ID>)
5.1.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
25 of 73
Variable name
Description
Definition Type
Class
Unit
Size
filename
Name of L1b science
data file to be processed
N/A
char
I
N/A
pointer
aocs_data
Structure with processed
AOCS data
N/A
t_l1a_aoc
s
I
N/A
1
pvt_data
Processed PVT data
N/A
t_l1a_pvt
I
N/A
1
l1b_data
L1b data
Section
3.2.2
t_l1b_bt_f I
req
N/A
1
earth_grid
Earth Fixed Grid
Section
3.2.1
double
I
N/A
4x2.6x10
^6
rfi_mask
RFI mask, 0- no RFI, 1 –
RFI present
Section
3.2.1
bool
I
N/A
2.6x10^6
earth_af_fov
AF-FOV border
projection into Earth grid
coordinates
Section
3.2.1
t_l1c_af_f
ov_earth
L
N/A
Scene
dependan
t
earth_pixel_list
List of pixels inside AFFOV border projection
into Earth grid
coordinates
Section
3.2.1
t_l1c_af_f
ov_list
L
N/A
Scene
dependan
t
u_v
U/V coordinates of nonredundant baselines
Section
7.2.4
t_l1c_bas
elines
L
N/A
NON_R
ED_BAS
ELINES
antenna_patt
Averaged LICEF receiver Section
directional power Gain
3.2.3
double_co I
mplex
N/A
1
directivity
Averaged directivity of
the LICEFs
Section
3.2.3
double
I
N/A
1
sys_temp
Averaged System
Temperature measured
by the PMS system
[AD.8]
double
I
Kelvin
1
l1c_data
L1c data per pixel
Section 4
t_l1c_data O
Kelvin
1
bt_accuracy
Radiometric Accuracy
Eq. 15
double
O
Kelvin
1
ionospheric_rotation_a
ngles
Ionospheric Rotation
Angles
Eq. 1
double
O
degrees
2
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
26 of 73
Variable name
Description
Definition Type
Class
Unit
Size
footprint_size
Equivalent Array Factor
Eq. 16
double
O
meters
2
pixel_observation_angl
es
Pixel observation angles
Section
3.1.3
t_l1c_obs
_angles
O
degrees
2
Table 4: Geolocation module variable list
5.1.4. Implementation
The geolocation processing shall ingest a L1b product file, and for each scene contained inside it shall
compute a list of Earth Fixed pixels observed, along with the magnitudes being measured (Brightness
Temperature, radiometric accuracy, observation angles, rotation angles, footprint size and quality flags)
The process, as indicated on figure 9, must first perform a sequence of computations valid for each
scene, and later on another sequence of computations for each pixel contained inside the scene. Details
are given on each sub-process in the next sections, using one section per box in figure 9.
Basically, the approach to be taken is to process each L1b scene at a time, computing the Earth Fixed
pixels contained inside, and then for each pixel compute the corresponding output values. In the nominal
case, most of the pixels computed for one scene should have been also computed in the previous one,
with the exception of some new ones appearing and others no longer visible as the S/C moves between
scenes. The end result shall be a list of pixels (around 80000 different ones for a 2500 scene product),
and for each pixel there shall be a list of associated measurements, taken as it traverses the FOV from
the top to the bottom.
What shall be described here is the consolidation approach to be taken, in order to store in L1c products
only pixels whose measurement lists are complete.
This can be achieved by first limiting the output of the product by taking out all pixels contained in the
last scene set in the L1b product. This ensures that all pixels to be written in the L1c product shall
contain the whole list of measurements (i.e. observation angles). A good measure is to control this
behaviour through a configuration flag in the L1c processing.
A second step needed is to process this L1b product together with the previous L1b product. This allows
introducing in the product the pixels belonging to the last scene of the previous L1b product, and all
their measurements, which shall complement the measurements available from the current L1b product
processing.
In order to achieve this, the processing of this previous product must be performed “backwards”,
starting from the last L1b scene, and then progressing to previous scenes, but only using the pixels
filtered from the last scene. This will insert into the L1c product the pixels and measurements that were
removed when the previous product was processed.
A final step is to extract the brightness temperatures for pixels with an observation angle of 42.5º and
create a browse field, also part of the l1c_data product.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
27 of 73
5.1.5. Error Handling
Error Nr Error Description Error Code Returned
1
File accessor
unable to read the
L1b file
DATA_CORRUPTED
Program behaviour
Quality flag
raised
Processing unit returns with
no data processed.
No
Table 5: Geolocation module error handling
5.2. Initialise Attitude
This computation initialises all the EE CFI required structures that shall be used in later functions. These
structures are composed by a time_id and an attitude_id structure.
The time_id shall be used in all time conversions and coordinate system change. It can be initialised at
the beginning of processing. However, the attitude_id structure must be computed for each scene, as it
represents the pointing angles for the instrument at a particular time.
5.2.1. Inputs
L1b data (SM_xxxx_ MIR_SC_D1B<ID> or SM_xxxx_ MIR_SC_F1B<ID>)
•
UTC time
•
PVT and AOCS
Auxiliary Data
•
Time Correlations (SM_xxxx_MPL_ORBSCT<ID>)
•
Bulletin B ((SM_xxxx_AUX_BULL_B_<ID>)
•
Best Fit Plane (SM_xxxx_AUX_BFP___<ID>)
•
Mispointing Angles (SM_xxxx_AUX_MISP__<ID>)
5.2.2. Outputs
EE CFI structure time_id
EE CFI structure attitude_id
matrix_BFP2EF
5.2.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
28 of 73
Variable name
Description
Definition Type
Class Unit
Size
pvt_data
Processed PVT data
N/A
t_l1a_pvt
I
N/A
1
utc_time
UTC snapshot time
N/A
double
I
days
1
aocs_data
Processed AOCS data
N/A
t_l1a_aocs
I
N/A
1
BF_to_BFP_matrix
Body Frame to Best Fit
Plane rotation matrix
N/A
double
I
N/A
3x3
STR_to_BF_matrix
Star Tracker to Body
Frame rotation matrix
N/A
double
I
N/A
3x3
time_id
EE CFI time structure
N/A
xl_time_id
O
N/A
1
attitude_id
EE CFI pointing structure N/A
xp_attitude O
_id
N/A
1
matrix_BFP2EF
Rotation matrix between N/A
the BFP and the EF
frame
double
N/A
9
O
Table 6: Initialise Attitude variable list
5.2.4. Implementation
First of all, the time_id structure must be initialised using time correlations available in any reference
file. Currently the most recent IERS bulletin B is used as input auxiliary file, but in case it is not
available, an older bulletin B or even the Orbit Reference File can be used to retrieve these values.
Once this correlations are obtained, the time_id structure is initialised using the EE CFI function
xl_time_ref_init using as input those correlations and receiving the time_id structure.
Once a valid time_id structure is available, the attitude_id must be initialised using the scene time, PVT
and AOCS data. The initialisation is done using EE CFI function:
xp_attitude_user_set(&time_id, attitude_id, /* input / output */
&time_reference, &time, pvt_data.position, pvt_data.velocity,
acceleration, &target_frame, TOD_to_BFP_matrix, matrix_rate, matrix_rate_rate,
offset, ierr_xp)
The inputs for this function are the following:
1. time_id (previously computed)
2. time_reference (set to XL_TIME_UTC)
3. time is the scene time expressed in decimal days since 1st Jan 2000 00:00:00
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
29 of 73
4. pvt_data.position
5. pvt_data.velocity
6. acceleration (set to [0, 0, 0])
7. target_frame (set to XP_INSTR_ATT)
8. TOD_to_BFP_matrix (computed in next step)
9. matrix_rate (set to [0, 0, 0, 0, 0, 0, 0, 0, 0])
10. matrix_rate_rate (set to [0, 0, 0, 0, 0, 0, 0, 0, 0])
11. offset (set to [0, 0, 0])
The output shall be the attitude_id structure already initialised.
The matrix TOD_to_BFP_matrix (True Of Date to Best Fit Plane) shall be computed by the
multiplication of three other matrices: STR_BF_matrix (Star Tracker to Body Frame), BF_BFP_matrix
(Body Frame to Best Fit Plane) and TOD_to_BF_matrix (True Of Date to Body Frame), described
below.
The BF to BFP rotation matrix takes in account the antenna phase centres of the MIRAS array, and
corresponds to a rotation from the Body Frame with Euler angles yaw (Ψ), pitch (Θ), and roll (Φ) to the
Best Fit Plane. These values are taken from the Best Fit Plane ADF by combining the two rotations
contained inside this file (on-ground measured BFP rotation and in-orbit calibrated BFP rotation).
On an equivalent approach, the STR to BF rotation matrix takes into account the possible mispointing
that can be accumulated between the Star Tracker and the Antenna frame due to deformations of the
payload structure. At launch it will be initialised to zero, but may be modified in flight if such
deformations are detected. This STR to BF rotation also corresponds to a rotation from the Star Tracker
Frame with Euler angles yaw (Ψ), pitch (Θ), and roll (Φ) to the Body Frame. These values are taken
from the Mispointing ADF by combining the three rotations contained inside this file (Proteus to Body
rotation, Body to Antenna rotation and in-orbit calibrated Mispointing rotation).
Both matrices are constructed by multiplying, in the inverse order, the rotation matrices for each Euler
angle, resulting in the following matrix:
cos Φ cos Θ cos Φ sin Θ cos Ψ + sin Φ sin Ψ cos Φ sin Θ sin Ψ − sin Φ cos Ψ 
 − sin Θ

cos Θ cos Ψ
cos Θ sin Ψ


 sin Φ cos Θ sin Φ sin Θ cos Ψ − cos Φ sin Ψ sin Φ sin Θ sin Ψ + cos Φ cos Ψ 
Eq. 32
In order to compute the matrix TOD_to_BF_matrix (True Of Date to Body Frame), two other matrices
must be computed and multiplied.
First the matrix J2000 to Body Frame is computed, by converting the quaternions in the aocs_data
structure into vectors. This is done by using function:
xl_quaternions_to_vectors (quaternions, ux_vec, uy_vec, uz_vec, ierr_xl)
Special care must be taken with the quaternion ordering, as for SMOS data the real element is the first
element, whereas in this CFI function it must be the last. The resulting vectors set in a matrix and with
negative sign represent the J2000 to BF rotation matrix. The negative sign is needed due to different
conventions between SMOS and CFI functions
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
30 of 73
The second matrix to be computed is the TOD to J2000, which is obtained by transforming the three
unitary vectors from TOD to GM2000 using CFI function:
xl_change_cart_cs(&time_id, &mode, &cs_in, &cs_out, &time_reference, &time,
pos_in, vel, acc, pos_out, vel_out, acc_out)
Where the input variables are:
1. cs_in (set to XL_TOD)
2. cs_out (set to XL_GM2000)
3. pos_in (set alternatively to [1, 0, 0], [0, 1, 0] and [0, 0, 1])
4. vel and acc (set to [0, 0, 0])
The three pos_out vectors, set in matrix form, compose the TOD to J2000 rotation matrix.
These four matrices must be now multiplied in order, and as a final step, in the resulting matrix the first
line must be swapped with the last line in order to cope with the different axis convention between
SMOS and CFI.
 X SMOS = − Z CFI

YSMOS = −YCFI
Z
 SMOS = − X CFI
Eq. 33
Additionally, from the TOD_to_BF_matrix a new rotation matrix is created (matrix_BFP2EF) using the
same CFI function as before but with input variables:
1. cs_in (set to TRUE_OF_DATE)
2. cs_out (set to EARTH_FIXED)
3. pos_in (columns of ToD_to_BF_matrix)
4. vel and acc (set to [0, 0, 0])
This rotation matrix is used in other routines of L1PP, such as the Pixel Footprint calculation (Section
5.11)
5.2.5. Error Handling
In case of time_id initialisation failure, the processing of the L1b product must be aborted. There can be
no time conversions without a valid time_id.
In case of attitude_id initialisation failure, the affected scene must be skipped, but the process of the
remaining scenes can continue. There cannot be any pixel geolocation without a correct attitude_id valid
for each scene.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
31 of 73
5.3. Ionospheric Correction module
Figure 12: Ionospheric Correction module processing steps
The Ionospheric Correction module computes the geomagnetic field parameters based on the
International Geomagnetic Reference Field model (IGRF2010). The Total Electron Content is also
computed, based on Total Electron Content (TEC) ADF maps or on the default IRI2001 model.
5.3.1. Inputs
L1b Data (SM_xxxx_MIR_SC_D1B_<ID> or SM_xxxx_MIR_SC_F1B_<ID>)
•
UTC time
•
PVT Data
Auxiliary Data
•
Vertical Total Electron Content Data (SM_xxxx_AUX_VTECy_<ID>, with y being P/R/C)
•
Geomagnetic Model Data file available with the IGRF 11th generation model
5.3.2. Outputs
Total Electron Content Data (TEC)
Magnetic Field Strength (F)
Magnetic Field inclination (I) and declination (D)
5.3.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
sc_position
Processed PVT data
N/A
double
I
m
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
3
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
32 of 73
Variable name
Description
Definition
Type
Class
Unit
Size
utc_time
UTC snapshot time
N/A
double
I
days
1
geo_results
Vector to hold full N/A
IGRF2010 model output
double
L
N/A
8
dyear
UTC snapshot time
expressed in decimal
years
N/A
double
I
year
1
x_lat
IRI2001/ IGRF2010
interface, Satellite
latitude
Geodetic
latitude
double
L
degrees
1
x_long
IRI2001/ IGRF2010
interface , Satellite
longitude
Geodetic
longitude
double
L
degrees
1
x_height
IRI2001/IGRF2010
interface, Satellite
altitude
Altitude
double
L
m
1
jf
IRI2001 interface, model
options
IRI2001
function
header
int
L
N/A
30
jmag
IRI2001
interface, IRI2001
optional input parameters function
header
int
L
N/A
1
year
IRI2001 interface, integer IRI2001
year from UTC time
function
header
int
L
years
1
mmdd
IRI2001 interface, integer IRI2001
month and day from UTC function
time (MMDD)
header
int
L
N/A
1
hour
IRI2001
interface, IRI2001
decimal hour from UTC function
time
header
float
L
hours
1
htecmin
IRI2001 interface,
minimum height for TEC
integration
IRI2001
function
header
int
L
km
1
geomag_f
Magnetic field strength
Section
3.1.1
double
O
Tesla
1
geomag_d
Magnetic
declination
field Section
3.1.1
double
O
degrees
1
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Variable name
Description
geomag_i
tec
Definition
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
33 of 73
Type
Class
Unit
Size
Magnetic field inclination Section
3.1.1
double
O
degrees
1
Total Electron Content
double
O
TEC
1
Section
3.1.1
Table 7: Ionospheric Correction variable list
5.3.4. Implementation
The geomagnetic model is available from the Fortran model IGRF2010 model, as explained in Section
3.1.1, and a suitable C wrapper must be implemented to interface it. First, the relevant satellite data must
be computed from the UTC time and PVT data retrieved for every snapshot processed:
1. UTC time shall be converted from its input format into decimal years
2. PVT position data shall be converted to geodetic latitude and longitude using CFI function
xl_cart_to_geod
The
IGRF2010
baseline
model
shall
be
the
C
model
available
in
st
http://www.ngdc.noaa.gov/IAGA/vmod/igrf.html, and available at 31 March 2010. This C model
contains more expansion coefficients beyond 2010 than the fortran counterpart.
The IGRF2010 model requires these three inputs above (dyear, x_lat and x_long), plus the altitude at
which the geomagnetic field will be computed. In this case, the altitude desired (x_height) shall always
be 450km:
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
double geo_results[8];
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
34 of 73
// Vector to hold full Igrf2000 model output
//Compute the geomagnetic field, based in the International Geomagnetic
//Reference Field model. It is implemented in a set of C functions, obtained
//from the IAGA site. These can be found at
//http://www.ngdc.noaa.gov/IAGA/vmod/igrf.html
error = computeIgrf2000(dyear, x_height, x_lat, x_long, geo_results[0],
geo_results[2],
geo_results[3],
geo_results[4],
geo_results[5],
geo_results[7]);
geo_results[1],
geo_results[6],
If(error != 0)
Error #1 handled
// We are only interested three parameters from the eight returned by Igrf2000 routine
geomag_f = geo_results[0];
// babs field in Igrf2000 interface
geomag_i = geo_results[4];
// dip field in Igrf2000 interface
geomag_d = geo_results[5];
// dec field in Igrf2000 interface
The resulting variables are geomag_f (magnetic field strength), geomag_d (magnetic field declination),
and geomag_i (magnetic field inclination), all values at satellite height.
Next, the TEC value applicable to that position and time can be computed from the available ADF by
interpolation, or using the IRI2001 model in case there are no files available:
In case of interpolation using ADF, the files used are the IGS combined, which contain data on a
mercator grid and a temporal sampling of 2 hours. Each file contains 13 layers, being the first layer a
repetition of the last layer from the previous file.
Thus, the interpolation shall be first performed as a spline (cubic) interpolation in time, taking into
account the UTC time requested and the UTC times of the datasets (e.g. possibly generating an
intermediate layer over the whole mercator grid). The spline takes into account the 13 layers used, so
that the complete behaviour of the TEC field is captured over 24h.
Afterwards, a simple bi-linear interpolation is required in lat-lon (interpolating between the four closest
grid points for each lat-lon point). A note of advice is mentioned here, as the data in the IGS combined
is expressed in 0.1 TEC Units.
If the option to compute the TEC value using the IRI2001 model is selected, then this FORTRAN model
(in particular function irit13ne) must be invoked using the following parameters:
1. x_lat and x_long as computed before from PVT position data
2. jmag option as 0.
3. jf model options vector as all 1, except for positions 5, 6 and 21 which should be 0.
4. UTC time converted into Local Time and expressed as integer year, integer month+day (MMDD)
and decimal hour (conversion from UTC into Local Time is achieved by adding 25 hours to the
decimal hours)
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
35 of 73
5. start and stop altitudes, where the start altitude is selected as 50km and stop altitude is 450km.
The output of this function shall be the TEC value expressed in TEC Units (1 TECU = 1e+016e-/m2)
The
IRI2001
baseline
model
shall
be
the
fortran
model
available
in
st
ftp://nssdcftp.gsfc.nasa.gov/models/ionospheric/iri/iri2001/fortran_code, and available at 31
October 2007.
In either case, the resulting value, tec, is the Total Electron Content for all pixels in the snapshot. The
module shall return geo_f, geo_d, geo_i and tec as output for the input UTC time and PVT position.
5.3.5. Error Handling
Error Nr Error Description Error Code Returned
Program behaviour
Quality flag
raised
1
IGRF2010 model
returns error
DATA_CORRUPTED
Processing unit sets default
values for geomagnetic data
and continues.
Yes
2
File accessor
unable to read the
VTEC AUX file
DATA_CORRUPTED
Processing unit sets default
values for TEC data and
returns. Use of IRI2001
model should be tried next
Yes
3
IRI2001 model
returns error
DATA_CORRUPTED
Processing unit sets default
values for TEC data and
returns.
Yes
Table 8: Ionospheric Correction module error handling
5.4. Compute EAF-FOV
This algorithm computes for each scene a discrete list of xi-eta points in the antenna frame, depicting
the Extended Alias-Free FOV. It shall also compute two important angles for the spacecraft, which shall
be used in further computations: the tilt angle and the local_north angle.
The first computation of the EAF-FOV shall make use of the initialised attitude_id structure only,
whereas the angles computation shall also use the pvt_data structure from each scene, containing
position and velocity.
5.4.1. Inputs
attitude_id
pvt_data
5.4.2. Outputs
eaf_fov_list
tilt_angle
local_north_angle
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
36 of 73
5.4.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
pvt_data
Processed PVT data
N/A
attitude_id
EE CFI pointing structure Section
5.2.3
eaf_fov_list
List of xi-eta points in the N/A
border of the EAF-FOV
tilt_angle
Angle between S/C to N/A
Nadir
direction
and
Instrument boresight
local_north_angle
Angle
between
S/C N/A
ground
track
instantaneous direction
and local north
t_l1a_pv
t
I
N/A
1
xp_attitu
de_id
I
N/A
1
O
N/A
1
double
O
deg
1
double
O
deg
1
Table 9: Compute EAF-FOV variable list
5.4.4. Implementation
The computation of the EAF-FOV is done through the selection of a geometrical boundary that follows
certain requirements. These requirements can be observed in figure 6, in section 3.21.
The area must be limited in the first place by a hexagon of radius
2
(where d is 0.875), and also by the
3d
Earth aliases in all six adjacent aliases.
Thus, one of the first tasks is to obtain the discrete contour of the Earth within the unit circle. This
contour is obtained by projecting the azimuth and elevation coordinates of the Earth tangent points into
the cosines domain. This step needs to be done using EE CFI functions, and is composed of the
following steps:
1. Establish the discrete list of azimuth directions, for example from 0º to 360º in incremental steps of
1º
2. For each azimuth direction, call xp_target_altitude using attitude_id structure and the azimuth as
input, in order to obtain an initialised target_id structure. Parameter dem_id and atmos_id shall be
left NULL, deriv shall be XP_NO_DER.
3. Call xp_target_extra_main with the target_id structure obtained in the previous step, select choice
XP_TARG_EXTRA_MAIN_SAT2TARG_ATTITUDE,
target_number
shall
be
the
num_user_target returned from the previous function and target_type XP_LOS_TARGET_TYPE, in
order to obtain several output vectors (main_results, main_results_rate, main_results_rate_rate)
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
37 of 73
4. The azimuth angle in attitude frame is contained in main_results[6] (φ), and the elevation in
main_results[7] (θ). Taking into account the CFI angle convention, the cosines domain xi-eta
coordinates are computed according to equation 7
5. Gather all xi-eta pairs into a list (eaf_fov_list) to be used later
In order to shift these coordinates to each of the aliases, it must be known that the centers of the aliases
are located at:

cos (α ) 
π π 5π 7π 3π 11π
d 3

, ,
 with α = , , ,
2
6 2 6 6 2 6

eta =
sin (α )

d 3
xi =
2
Eq. 34
With all these inputs known, the selection of the contour passes by establishing the geometrical
requirements in an ad-hoc algorithm. For example, the algorithm implemented in L1PP travels trough a
discrete list of xi-eta points of the hexagon, and for any given xi value, it selects the appropriate eta
value from between the eta value of the hexagon and of the six aliases.
Additionally, the other two angles needed as output are obtained through a series of calls to EE CFI
functions. The only inputs required are the attitude_id and the pvt_data structure.
The processing steps are the following:
1. Call xl_cart_to_geod to transform the pvt_data into subsatellite point lat-lon-alt coordinates
2. Use the previous lat-lon coordinates, together with altitude 0 in function xl_geod_to_cart to obtain
the position and velocity coordinates of the subsatellite point.
3. Call xp_target_generic with the attitude_id structure plus the position and velocity values of the
subsatellite point, in order to obtain an initialised target_id structure. Parameter dem_id shall be left
NULL, deriv XP_NO_DER
4. Call xp_target_extra_main with the target_id structure obtained in the previous step, select choice
XP_TARG_EXTRA_MAIN_SAT2TARG_ATTITUDE,
target_number
shall
be
the
num_user_target returned from the previous function and target_type XP_LOS_TARGET_TYPE, in
order to obtain several output vectors (main_results, main_results_rate, main_results_rate_rate)
5. The tilt angle is computed from element number 8 in the main_results vector (tilt = 90 –
main_results[7])
6. Call xp_target_extra_aux with the target_id structure obtained in the step 3, select choice
XP_TARG_EXTRA_AUX_TARGET_NADIR_VEL, target_number shall be the num_user_target
returned from the previous function and target_type XP_LOS_TARGET_TYPE, in order to obtain
several output vectors (aux_results, aux_results_rate, aux_results_rate_rate)
7. The local north angle is computed from element number 11 in the aux_results vector
(aux_results[10]), which represents the azimuth of nadir velocity with respect to local north
5.4.5. Error Handling
In case any error happens in the contour computaion, the processing of the related scene must be
aborted. Without the xi-eta contour the DGG list of points cannot be extracted.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
38 of 73
In case the error happens during the computation of the tilt, the geometrical rotation of all pixels in that
scene cannot be computed.
In case the error happens during the computation of the local north angle, the geometrical rotation and
the footprint of the pixels cannot be computed.
5.5. Retrieve DGG pixel list
This algorithm shall filter out the list of Earth fixed pixels available from the ADF according to the
contour computed in the previous step. In order to this, first the points in the contour list must be
projected onto the Earth’s surface. Afterwards, all pixels in the ADF shall be compared against the
contour and only those falling INSIDE the contour shall be selected for the scene.
It is advisable at this stage to ingest not only the DGG ADF, but also the Land-Sea mask and the RFI
mask, as it is a good point to associate each of the final selected pixels to their corresponding flags.
5.5.1. Inputs
attitude_id
eaf_fov_list
Auxiliary Data
•
Earth Fixed Grid (SM_xxxx_AUX_DGG___<ID>)
•
Land Sea Mask (SM_xxxx_AUX_MASK__<ID>)
•
RFI mask (SM_xxxx_AUX_RFI___<ID>)
5.5.2. Outputs
earth_pixel_list
5.5.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
attitude_id
EE CFI pointing structure Section
5.2.3
eaf_fov_list
earth_pixel_list
xp_attitu
de_id
I
N/A
1
List of xi-eta points in the Section
border of the EAF-FOV
5.4.3
I
N/A
1
List of pixels from the N/A
DGG ADF contained
inside the EAF-FOV
border
O
N/A
1
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
39 of 73
Table 10: Compute DGG pixels variable list
5.5.4. Implementation
First of all, each of the xi-eta points in the contour list of the EAF-FOV must be projected onto the
Earth. This is done through the use of several EE CFI functions as in the following approach:
1. For each xi-eta point extract the corresponding azimuth and elevation angles. Azimuth must be
converted to [0,360] range before being used by CFI functions.
2. Call xp_target_inter function with attitude_id structure, the azimuth and elevation angles, deriv
XP_NO_DER and inter_flag XP_INTER_1ST, in order to obtain an initialised target_id structure.
Parameter dem_id and atmos_id shall be left NULL.
3. Call xp_target_extra_main with the target_id structure obtained in the previous step, select choice
XP_TARG_EXTRA_MAIN_GEO, target_number shall be the num_user_target returned from the
previous function and target_type XP_LOS_TARGET_TYPE, in order to obtain several output
vectors (main_results, main_results_rate, main_results_rate_rate)
4. The main_results vector contains the geodetic latitude and altitude, as well as geodetic and
geocentric longitude.
5. The geodetic latitude and altitude and the geocentric longitude shall be used in another call to
xl_geod_to_cart function, in order to obtain the Earth Fixed Cartesian coordinates (X, Y Z) of that
particular point.
6. Both Earth Fixed and geodetic/geocentric coordinates should be stored for later usage.
Once the EAF-FOV border has been projected into Earth Fixed coordinates, it must be used as a filter
on all the points available in the DGG ADF.
As a first filter, the bounding box of the contour in lat-lon coordinates can be easily computed with the
maximum and minimum values of latitude and longitude for all points in the contour. This guarantees
that not all pixels from the DGG shall be passed to the second filter.
The second filter needs to be a bit more sophisticated, in order to take into account the strange geometry
of the contour once it has been projected. This is done, for example, by taking the longitude of the pixel
that needs verification of whether it is inside or outside the zone, and computing the intersections of this
constant longitude with the contour. There can be several cases:
No intersection: Means that the pixel is outside the zone
Intersection in one point only: Means that the pixel outside the zone, or that one pole is inside the
zone. If the pole is inside the zone, then if the latitude of the pixel is between the pole and the
latitude of the intersection, the pixel is inside, otherwise it is outside
Intersection on two points: If the latitude of the pixel is within the latitudes of the intersections, it is
inside, otherwise it is outside
Intersections on more than two points are expanded cases of the ones presented before.
Once the list of Earth Fixed pixels has been extracted, it is also interesting to use their latitude, longitude
and altitude values to obtain the Earth Fixed Cartesian coordinates. This is done using the
xl_geod_to_cart function described in previous points.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
40 of 73
It is important to create at this point the distinction between land and sea pixels, according to the
AUX_MASK ADF, which for each unique pixel id in the DGG contains a land and a sea flag for each
id. It must be mentioned, that one pixel may have one or both flags at the same time (i.e. coastlines are
to be processed as land and sea)
It should also be a good place to extract the RFI flag for the pixels as well, which will be later used in
the flagging of measurements.
5.5.5. Error Handling
In case any error happens during projection of the EAF-FOV or extraction of the Earth Fixed pixels, the
processing of the scene must be aborted. The other scenes can continue to be processed nominally.
5.6. Compute Pixel Angles
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to
compute all the relevant angles from S/C to pixel in several reference frames.
5.6.1. Inputs
attitude_id
earth_pixel (taken from earth_pixel_list)
5.6.2. Outputs
satellite_to_target_sc_frame (azimuth and elevation)
satellite_to_target_topo_frame (azimuth and elevation)
target_to_satellite_topo_frame (azimuth and elevation)
5.6.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
attitude_id
EE CFI pointing structure Section
5.2.3
earth_pixel
Pixel extracted from the Section
list of pixels contained 5.5.3
inside the EAF-FOV
border
satellite_to_target_sc_f
rame
S/C to pixel directions N/A
expressed in the S/C
frame
xp_attitu
de_id
double
I
N/A
1
I
N/A
1
O
deg
2
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Variable name
Description
satellite_to_target_topo
_frame
target_to_satellite_topo
_frame
Definition
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
41 of 73
Type
Class Unit
Size
S/C to pixel directions N/A
expressed in topocentric
frame
double
O
deg
2
Pixel to S/C directions N/A
expressed in topocentric
frame
double
O
deg
2
Table 11: Compute Pixel Angles variable list
5.6.4. Implementation
Based on the pixel Earth Fixed Cartesian coordinates, a selection of EE CFI functions must be called in
order to compute the respective angles. The procedure is the following:
1. Call xp_target_generic with attitude_id structure, the EF position of the pixel, in order to obtain an
initialised target_id structure. Parameter dem_id shall be left NULL, deriv shall be XP_NO_DER
2. Call xp_target_extra_main with the target_id structure obtained in the previous step, select choice
XP_TARG_EXTRA_MAIN_ALL, target_number shall be the num_user_target returned from the
previous function and target_type XP_LOS_TARGET_TYPE, in order to obtain several output
vectors (main_results, main_results_rate, main_results_rate_rate)
3. The results are gathered from the main_results vector in the following way:
3.1. target_to_satellite_topo_frame[0] = main_results[8];/*azimuth target to S/C Topocentric CS*/
3.2. target_to_satellite_topo_frame[1] = 90.0 - main_results[9];/*elevation target to S/C Topocentric
CS, using complementary to set 0º on the pixel local vertical*/
3.3. satellite_to_target_topo_frame[0] = main_results[4];/*azimuth S/C to target Topocentric CS*/
3.4. satellite_to_target_topo_frame[1] = main_results[5];/*elevation S/C to target Topocentric CS*/
3.5. satellite_to_target_sc_frame[0] = main_results[6];/*azimuth S/C to target Attitude CS*/
3.6. satellite_to_target_sc_frame[1] = main_results[7];/*elevation S/C to target Attitude CS*/
5.6.5. Error Handling
In case of error during the angle computation, the specific pixel processing must be aborted. The rest of
the pixels in the scene can continue to be computed with no adverse effects. There is no need to abort
the scene processing, as only one pixel has been affected.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
42 of 73
5.7. Apodisation Window Computation
Figure 13: Apodisation Window module processing steps
The apodisation window computation will return the value of apodisation window for each pixel, based
on the antenna frame coordinates. Optionally, it will compute the strip adaptive apodisation window
term using also the (xi, eta) coordinates. The L1 Processor Prototype shall implement both apodisation
methods, although it shall be considered that the baseline for the L1 Operational Processor shall not
include Strip Adaptive processing.
5.7.1. Inputs
Antenna plane coordinates (xi,eta), non-redundant baseline coordinates (u,v)
attitude_id
earth_pixel (taken from earth_pixel_list)
Auxiliary Data
•
Apodisation
function
model
coefficients
(SM_xxxx_AUX_APDL__<ID>,
_xxxx_AUX_APDS__<ID> or _xxxx_AUX_APDNRT<ID>)
5.7.2. Outputs
Apodisation function term, W
5.7.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
43 of 73
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class
Unit
Size
xi_eta_pixel
Pixel xi-eta coordinates
(antenna frame)
Eq. 13
double
I
N/A
2
attitude_id
EE CFI pointing structure Section
5.2.3
xp_attitud
e_id
I
N/A
1
earth_pixel
Pixel extracted from the
list of pixels contained
inside the EAF-FOV
border
Section
5.5.3
I
N/A
1
apod_model1_points
Pre-computed cubic
coefficients for Strip
Adaptive Apodisation
model [log10(αu,/αv)]
Eq. 31
double
I
N/A
4
apod_model2_points
Pre-computed cubic
coefficients for Strip
Adaptive Apodisation
model [log10(αu,αv)]
Eq. 31
double
I
N/A
4
u_v
(u,v) coordinates of the
instrument baselines Set
of non-redundant
baselines (0 – redundant,
1 – non-redundant)
Section
7.2.4
t_l1c_bas
elines
I
N/A
NON_R
ED_BAS
ELINES
xi_eta_circle
xi-eta coordinates of 6
points in a circle around
the pixel projection
(antenna frame)
Eq. 7
double
L
N/A
2x6
w_max
Major semi-axis of
projected ellipse (antenna
frame)
Eq. 31
double
L
N/A
1
w_min
Minor semi-axis of
projected ellipse (antenna
frame)
Eq. 31
double
L
N/A
1
delta
Angle between major and
minor semi-axes
Eq. 28
double
L
N/A
1
alpha_u
Strip adaptive
Apodisation coefficient
Eq. 26
double
L
N/A
1
alpha_v
Strip adaptive
Apodisation coefficient
Eq. 26
double
L
N/A
1
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
44 of 73
Variable name
Description
Definition Type
Class
Unit
Size
w
Apodisation term
Eq. 26
O
N/A
NON_R
ED_BAS
ELINES
double
Table 12: Apodisation Window computation variable list
5.7.4. Implementation
If no strip adaptive method is selected, the Blackman Window is used to compute the apodisation factor
(Eq. 17):
//Compute w term using the Blackmann window
n_el=21;
//Nel for MIRAS instrument
d=0.875;
//Wavelengths between baselines
For(i=0... NON_RED_BASELINES)
double term_a = PI*(sqrt(u_v[i]->u* u_v[i]->u + u_v[i]->v* u_v[i]-> v)) / (sqrt(3) *
n_el *d);
w[i] = 0.42 + 0.5 * cos(term_a) + 0.08*cos(2*term_a);
EndFor
The rest of the section describes the strip adaptive method. This method was implemented for L1PP
v1.5, and although it is not to be considered for the L1 processing baseline, it has been kept in L1PP for
future analysis.
Based on the latitude and longitude coordinates of the earth_pixel, we shall compute 6 points over the
Earth’s surface located at a radius R from the pixel coordinates. This radius must be set as a
configuration value for the Strip Adaptive process, typically between 15-25km. Six separate points are
needed, as they shall be projected on the antenna frame and used to fit an ellipse.
Projection on the antenna frame shall be done following the same approach as described in section 5.6.4,
except that now we are only interested on getting the S/C to Target angles, so we can select as choice
XP_TARG_EXTRA_MAIN_SAT2TARG_ATTITUDE. The angles are transformed into xi-eta
coordinates using Eq.13.
With the collection of six xi-eta points, they can be used in any least squares method to fit the 6
parameters of an ellipse quadratic equation, for example using the method described in [RA.02].
Once the ellipse equation is known, obtaining both semi-axes and the delta angle is trivial.
We then proceed to combine these ellipse parameters with the pre-computed apodisation model
coefficients using the system of equations in Eq.31, and find the optimal parameters to use in Eq. 26:
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
45 of 73
//Eq. 25 of DPML1c
xx = log10(w_max * w_min); x = log10(w_max / w_min);
q_xx
=
apod_model2_points[0]*xx*xx*xx
apod_model2_points[2]*xx + apod_model2_points[3];
+
apod_model2_points[1]*xx*xx
+
p_x = apod_model1_points[0]*x*x*x + apod_model1_points[1]*x*x + apod_model1_points[2]*x +
apod_model1_points[3];
alpha_u = pow(10, (p_x + q_xx) / 2); alpha_v = pow(10, (q_xx - p_x) / 2);
//Compute Apodisation Window (Eq. 20 from DPML1c)
max_r=sqrt(3)*n_el*d;
For(i=0... NON_RED_BASELINES)
// Eq. 22 of DPML1c
rotateDirections(u_v[i]->u, u_v[i]->v, delta, &u_rot, &v_rot);
// Eq. 21 of DPML1c
double z_u = (alpha_u * sqrt(1 - (u_rot/max_r)^2));
double z_v = (alpha_v * sqrt(1 - (v_rot/max_r)^2));
double numerator_u = 1.0;
double numerator_v = 1.0;
For (i=1...20)
numerator_u = numerator_u + ((z_u*.5)^(2*i))/((fact(i)^2));
numerator_v = numerator_v + ((z_v*.5)^(2*i))/((fact(i)^2));
EndFor
double denominator_u = 1.0;
double denominator_v = 1.0;
For (i=1...20)
denominator_u = denominator_u + ((alpha_u*.5)^(2*i))/((fact(i)^2));
denominator_v = denominator_v + ((alpha_v*.5)^(2*i))/((fact(i)^2));
EndFor
// Eq. 20 of DPML1c
w(count) = (numerator_ u*numerator_v)/(denominator_u*denominator_v);
EndFor
The return of the function is w, a vector with the 2791 Apodisation Window values.
5.7.5. Error Handling
Error Nr Error Description Error Code Returned
1
Error in
interpolation for
apodisation
parameters
DATA_CORRUPTED
Program behaviour
Quality flag
raised
The element is set to a
default value (i.e. Blackman
default).
Yes
Table 13: Apodisation Window computation error handling
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
46 of 73
5.8. Compute Brightness Temperature
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to
perform a Discrete Fourier Transform. It shall perform a 2D integration, combining the xi-eta
coordinates of the pixel, the Brightness Temperature Fourier components from the L1b product and the
apodisation window computed before.
5.8.1. Inputs
L1b data (SM_xxxx_ MIR_SC_D1B<ID> or SM_xxxx_ MIR_SC_F1B<ID>)
•
ˆ
Reconstructed TB frequencies at top of the atmosphere
•
Constant Earth BT subtracted if Gibbs 1 method was applied
satellite_to_target_sc_frame (azimuth and elevation)
Apodisation function term, W
5.8.2. Outputs
Pixel brightness temperature
5.8.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
scene_bt_frequencies_d Brightness Temperature Section
ual
Fourier components for 3.2.2
the whole scene
Double I
complex
N/A
1396
scene_bt_frequencies_f
ull
Brightness Temperature Section
Fourier components for 3.2.2
the whole scene
Double I
complex
N/A
2791
constant_earth_BT
Brightness Temperature Section
for a constant Earth 3.2.2
subtracted during L1b
processing
Double
I
K
1
satellite_to_target_sc_f
rame
S/C to pixel directions Section
expressed in the S/C 5.6.3
frame
double
I
deg
2
w
Apodisation term
double
I
N/A
1396
Eq. 26
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
Date
:
29/11/2012
Issue
:
2.11
Page
:
47 of 73
Variable name
Description
Definition
Type
u_v
Non-redundant baselines
coordinates
section
7.2.4
pixel_bt_measurement
Integrated
Brightness N/A
Temperature for this
scene at that particular
viewing direction
SO-DS-DME-L1PP-0009
Class
Unit
Size
t_l1c_ba I
selines
N/A
1396
Double O
complex
K
1
Table 14: Compute BT variable list
5.8.4. Implementation
First of all, it must be indicated that there are two processing approaches, depending if the L1b data is
coming from dual polarisation or full polarisation. The only difference lies in the quantity of L1b data
required, which is 1396 complex values for H or V scenes, and 2791 complex values for HV scenes.
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7:
The Brightness Temperature measurement for H and V scenes is computed by performing a Discrete
Fourier Transform, equivalent to integration over each of the non-redundant baseline coordinates. The
first baseline corresponds to the centre frequency, so it contains only a real value. The integration must
be performed over the remaining 1395 baselines, taking into account that the data available in each L1b
scene is only half of the star domain.
As it has been described in [AD.8], each Brightness Temperature Frequency component corresponds to
a specific u,v baseline. The integration is performed over one half of the star domain by multiplying the
1395 elements in the BT frequencies vector element by element with the apodisation window vector and
with an equivalent vector containing the exponential term. Then it must be completed with the other half
of the star domain, by simply changing the sign of the u,v baseline and doing the complex conjugate of
the BT frequencies:
T (ξ ,η ) =
3d 2
3d 2
2π i ( uξ + vη )
ˆ
T
u
,
v
⋅
W
u
,
v
⋅
e
+
(
)
(
)
∑
2 u ,v
2
∑ Tˆ ( u, v ) ⋅W ( u, v ) ⋅ e π ( ξ
*
2 i u + vη )
Eq. 35
− u ,− v
The apodisation window should always be symmetrical, so the same term can be used for the u,v
baseline or for the -u,-v baseline.
After computing the Brightness Temperature from the frequency components, the extended sources that
may have been subtracted during L1b processing, in order to perform Gibbs error mitigation, have to be
added back. Depending on the Gibbs level it can be a constant flat Earth (Gibbs=1), or a constant flat
Land plus a model Ocean (Gibbs=2). The flat Earth or Land values are annotated in the L1b product,
whereas the Ocean model is the same as described in section 2.2.2.1.4 of [AD.8]. The discrimination
between Land and Ocean in order to add back the constant Land or the model Ocean for Gibbs 2
baseline is done based on the pixel water fraction content, considering it as Land if the water fraction is
below 50% and Ocean if it is above or equal to 50%.
In full polarisation measurement, the H and V scenes are processed in the same way as above, whereas
the HV scenes are processed by combining the 2791 complex values available. In this case, there is no
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
48 of 73
need to compute the complex conjugate for the -u,-v baselines, as described in [AD.8], the HV BT
frequencies cover the whole star domain so the DFT can be performed in a straightforward manner.
In full polarisation scenes (HV pol) there is no Gibbs term to be added back, it is only applicable to dual
polarisation scenes.
5.8.5. Error Handling
No errors are expected, unless wrong numbers (i.e. NaN) are coming from previous computations or
from the L1b products. The type of computation required is done inside a for loop, with no possibility of
division by zero or any other mathematical error.
5.9. Compute Radiometric Accuracy
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to
extract an estimation of the Radiometric Accuracy for the Brightness Temperature computed before.
The accuracy for each measurement will vary according to the xi-eta coordinates it has.
Several auxiliary values are required here, based on equation 9.
5.9.1. Inputs
satellite_to_target_sc_frame (azimuth and elevation)
W is the apodisation window term for each (u,v) baseline
Auxiliary Data
•
Average Antenna Patterns (SM_xxxx_AUX_PATT<ID>)
G pq (ξ ,η ) is the averaged LICEF receiver directional power Gain function normalised
so that it is unity at boresight.
•
Default PLM values (SM_xxxx_AUX_PLM___<ID>)
f0 is the central frequency and f01 is the low frequency
D is the averaged directivity of the LICEFs for each polarisation.
L1a HKTM Data (SM_xxxx_TLM_MIRA1A_<ID>)
•
Tsyspq is the averaged System Temperature measured by the PMS system, which has been used
to de-normalise the L1a calibrated visibilities.
d is the distance ratio between receivers (0.875)
B is the equivalent receiving frequency bandwidth in Hz, currently being 19 MHz
τ eff is the effective integration time, and is equivalent to τ ceff where τ is the integration time (1.2s
for H and V scenes and 0.8s for HV scenes) and ceff is the coefficient that accounts for the 1-bit
correlation, oversampling and hermiticity (ceff=1.81).
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
49 of 73
5.9.2. Outputs
pixel_radiometric_accuracy
5.9.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
satellite_to_target_sc_f
rame
S/C to pixel directions Section
expressed in the S/C 5.6.3
frame
double
I
deg
2
w
Apodisation term
Eq. 26
double
I
N/A
1396
D
Antenna directivity
Section
3.2.3
double
I
N/A
1
Center_frequency
Central frequency of Section
operation
of
the 3.2.3
instrument
double
I
MHz
1
Low_frequency
Lower limit frequency of Section
operation
of
the 3.2.3
instrument
double
I
MHz
1
B
Frequency Bandwidth
Section
3.2.3
double
I
MHz
1
Avg_Tsys
System
average
Temperatures Section
3.2.3
double
I
K
1
Avg_Pattern
Antenna Pattern average Section
module
3.2.3
double
I
N/A
181x181
R
the redundancy level of Section
each (u,v) baseline (i.e. 3.2.3
number of times that the
baseline
has
been
measured, 1 for nonredundant
baselines,
greater than 1 for the rest)
double
L
N/A
1396
Double
O
K
1
pixel_radiometric_accu
racy
N/A
Table 15: Compute Radiometric Accuracy variable list
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
50 of 73
5.9.4. Implementation
As explained initially in section 3.2.3, the following equation shall be used to compute the radiometric
accuracy for a given viewing direction:
∆T pq (ξ ,η ) =
Ω pq 1 − ξ 2 − η 2 3 2
d
2
G pq (ξ ,η )
Tsyspq
B ⋅τ eff
α wα ol
Eq. 36
Where the different parameters are computed as follows:
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7:
4π
for the current LICEFs, where D
D
is the averaged directivity of the LICEFs. D is taken from the AUX_PLM_ ADF file for each
polarisation. For full polarisation, DHV is computed as DH DV .
Ω pq is the solid angle of the antenna, and can be computed as
G pq (ξ ,η ) is the averaged LICEF receiver directional power gain function normalised so that it is
unity at boresight. This data shall be retrieved from the average antenna pattern ADF
(AUX_PATT__) for each polarisation, and the value corresponding to the exact xi-eta coordinates
extracted by interpolation. For full polarisation, G HV (ξ ,η ) is computed as
G HH GVV .
d is the distance ratio between receivers (0.875)
Tsyspq is the averaged System Temperature measured by the PMS system, which has been used to de-
normalise the L1a calibrated visibilities. These values are taken from the corresponding L1a HKTM
product, for the ancillary data produced at the same OBET time as the scene being processed.
B is the equivalent receiving frequency bandwidth in Hz, currently being 19 MHz. B is taken from
the AUX_PLM_ ADF file.
τ eff is the effective integration time, and is equivalent to τ ceff where τ is the integration time and ceff
is the coefficient that accounts for the 1-bit correlation, oversampling and hermiticity (ceff=1.81). For
pure scenes (H and V pol) the integration time is 1.2s, for mixed scenes is 0.46s (H and V pol), and
for HV real and imaginary it is 0.64s (i.e. T3 and T4).
αw =
∑∑
u
v
(W ( u, v ) )
R ( u, v )
2
accounts for the apodisation window and redundancies in the
measurements, where W is the apodisation window term for each (u,v) baseline and R is the
redundancy level of that same baseline (i.e. number of times that the baseline has been measured, 1
for non-redundant baselines, greater than 1 for the rest)
 f −f 
−2π  0 01 
 B 
2
α ol = 1 + e
accounts for the local oscillator factor, where B is the bandwidth mentioned
before, f0 is the central frequency and f01 is the low frequency. Both frequency values f0 and f01 are
taken from the AUX_PLM_ ADF file
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
51 of 73
In order to compute the redundancy level R, the following computations must be performed, combining
real time data with auxiliary data:
First the LICEF failures (if any) must be retrieved from the L1a HKTM data whose OBET
corresponds to the OBET of the scene being processed. The default failures stored in the
AUX_FAIL__ ADF shall also be considered.
Secondly, the LICEF coordinates of all LICEFs shall be extracted from the AUX_PLM__ ADF,
together with the central frequency value f0
For each LICEF-LICEF pair, and in case they either one or the other are not failing, compute the u,v
baseline coordinates using the LICEF positions and central frequency. This will point to one unique
position in the star domain u_v vector, which should be incremented by one to reflect that the
baseline has been measured.
After running through all possible LICEF pairs, the output shall be a vector R, which for each u_v
position in the star domain will contain the number of times that the baseline was measured by the
instrument in that scene.
For the 0,0 baseline, the number of redundancies shall be the number of NIR measurements available
for that scene (3 if no failures, less in other cases).
5.9.5. Error Handling
In case applicable ADF files cannot be found, the processing must be aborted. The same case applies if
the average System Temperatures cannot be obtained for the corresponding OBET.
If the LICEF Failures cannot be obtained for the corresponding OBET, the default values can always be
used, but it is most probably that if the LICEF failures cannot be obtained for an OBET, most probably
the System Temperatures will not be obtained either, reverting to the initial case.
In case the LICEF failures are too many, it may be possible that some of the values in the Redundancy
vector are computed as zero. In this case, it is not advisable to use it in the equation, as it will cause a
division by zero exception. A warning should be issued, and a default value of one should be used, in
order to get a more reasonable value of the redundancy.
5.10. Compute Pixel Rotations
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to
extract the ionospheric and geometric polarisation rotation angles. These values are dependant on the
viewing angles, that is why they must be computed on a pixel basis.
Previously computed values are required here, as it has been shown in sections 3.1.3 and 3.1.4.
It should be noted that the implementation description has been expanded to include both methods,
although the baseline since L1PP v2.0 is only the Duesmann & Zundo implementation. The Waldteufel
& Caudal implementation is maintained for historical reasons.
5.10.1. Inputs
5.10.1.1. Waldteufel and Caudal Implementation
satellite_to_target_sc_frame (azimuth and elevation)
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
52 of 73
satellite_to_target_topo_frame (azimuth and elevation)
tilt angle
local north angle
Total Electron Content Data (TEC)
Magnetic Field Strength (F)
Magnetic Field inclination (I) and declination (D)
5.10.1.2. Duesmann and Zundo Implementation
satellite_to_target_sc_frame (azimuth and elevation)
satellite_to_target_topo_frame (azimuth and elevation)
satellite position in EF
rotation matrix between BFP and EF
Total Electron Content Data (TEC)
Magnetic Field Strength (F)
Magnetic Field inclination (I) and declination (D)
5.10.2. Outputs
Ionospheric_rotation_angle
Geometric_rotation_angle
5.10.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
satellite_to_target_sc_f
rame
S/C to pixel directions Section
expressed in the S/C 5.6.3
frame
Double
I
deg
2
satellite_to_target_topo
_frame
S/C to pixel directions Section
expressed in topocentric 5.6.3
frame
Double
I
deg
2
geomag_f
Magnetic field strength
Section
3.1.1
double
I
Tesla
1
geomag_d
Magnetic
declination
field Section
3.1.1
double
I
degrees
1
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Variable name
Description
geomag_i
Definition
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
53 of 73
Type
Class
Unit
Size
Magnetic field inclination Section
3.1.1
double
I
degrees
1
tec
Total Electron Content
double
I
TEC
1
tilt_angle
Angle between S/C to Section 5.4
Nadir
direction
and
Instrument boresight
double
I
deg
1
local_north_angle
Angle
between
S/C Section 5.4
ground
track
instantaneous direction
and local north
double
I
deg
1
satellite_position_EF
Position of the satellite in N/A
the EF frame
double
I
N/A
3
matrix_BFP2EF
Rotation matrix between Section 5.2
the BFP and the EF
frame.
double
I
N/A
9
Ionospheric_rotation_a
ngle
Faraday rotation angle
Section
3.1.3
double
O
degrees
1
Geometric_rotation_an
gle
Rotation angle due to Section
geometry
3.1.4
double
O
degrees
1
Section
3.1.1
Table 16: Compute Pixel Rotations variable list
5.10.4. Implementation
The Faraday rotation angle is easily computed once the TEC, geomagnetic field components and pixel
angles are available. Using equation 1, the following advice should be heeded:
All angles should be transformed to radians before using them inside the equation
TEC should be expressed in TECU
Geomagnetic field should be expressed in Tessla
θg φg are the satellite_to_target_topo_frame angles
ω = −6950 × F × TEC × sin I + cos I × tan θ g × cos (φg − D ) 
Eq. 37
It must be remarked that the value of θg above is not computed directly as the elevation in
satellite_to_target_topo_frame, as the EE CFI functions will return a negative angle. It has to be
computed as shown in Eq.40.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
54 of 73
Regarding φg, to be applied in Eq.37 it must be computed as obtained directly from the EE CFI, i.e.
satellite_to_target_topo_frame[0]. This is because in Eq.37 φg must be oriented in the same axis as D,
which is consistent with the EE CFI azimuth axis.
5.10.4.1. Waldteufel and Caudal Implementation
For the geometric rotation angle, an initial computation must be performed, in order to obtain the real
tilt angle of the array, as well as the angle of the sub-satellite track with the local north. These should
have been computed as indicated in section 5.4.
The equations shown in section 3.1.4 can be computed as follows:
θ=
π
φ =−
π
2
π
Eq. 38
+ satellite_to_target_sc_frame[0]
Eq. 39
+ satellite_to_target_topo_frame[1]
Eq. 40
− satellite_to_target_topo_frame[0]+local_north
Eq. 41
θg =
φg =
− satellite_to_target_sc_frame[1]
2
2
π
2
Depending on the magnitude of φg, the value of ϕ must be computed differently:
 cos(tilt_angle)sin θ g − sin(tilt_angle) cosθ g sin φ g
sin θ

 π
3π
  2 ≤ φg ≤ 2

 cos(tilt_angle)sin θ g − sin(tilt_angle) cos θ g sin φ g 
ϕ = π − arcsin 
 {otherwise
sin θ


ϕ = arcsin 
Eq. 42
The final geometric rotation angle must be computed as Geometric_rotation_angle α = φ – ϕ
5.10.4.2. Duesmann and Zundo Implementation
Setting
θe = target_to_satellite_topo_frame[1]
Eq. 43
ϕe = target_to_satellite_topo_frame[0]
Eq. 44
in for equations 7 and 8, the emission vectors are calculated in the topocentric frame.
With the CFI function XL_EF_TO_TOPOCENTRIC the rotation matrix from the BFP to EF is
transformed to the rotation matrix from the BFP to the pixel topocentric frame. Note that at this point,
the rotation matrix is written in CFI notation and it must be converted to SMOS notation, i.e.,
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
 X CFI = − X MIRAS

 YCFI = −YMIRAS
 Z =Z
MIRAS
 CFI
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
55 of 73
Eq. 45
This matrix allows the conversion of the emissions vectors in topocentric coordinates to MIRAS
coordinates.
With
θ s = satellite_to_target_sc_frame[1]
Eq. 46
ϕ s = satellite_to_target_sc_frame[0]
Eq. 47
being used to determine the vector from satellite to target (equation 9) the auxiliary variables are
computed and the Ludwig vectors computation is performed.
The geometric rotation angle is then taken from
r
r
 EhMIRAS ⋅ Ly MIRAS
r MIRAS
α ′ =arctan  r
 Eh
L
⋅
MIRAS
x





Eq. 48
5.10.5. Error Handling
In equation 42, depending on the values inside the arcsin function, there can be a NaN result if the
numerator is bigger than the denominator. In order to rule out limit cases within the tolerances, if the
ratio is slightly higher than 1, the value for ϕ shall be π/2, whereas if it is slightly smaller than –1, ϕ
shall be –π/2.
All other cases shall be treated as errors, and the computation of the affected pixel shall be skipped.
5.11. Compute Pixel Footprint
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to
extract the footprint semi-major and semi-minor axes, as it is approximated by an ellipse. The synthetic
beam shape of the antenna intersecting the antenna frame forms the footprint, once it is projected over
the Earth’s surface.
The common agreement is that the footprint is defined by the –3dB contour of the Equivalent Array
factor described in equation 10.
In order to avoid computing the Array Factor and projecting it for a lot of points and then looking for
the maximum and minimum values, it can be established that the maximum elongation of the ellipse is
achieved for the direction formed by the S/C nadir point and the pixel, whereas the minimum elongation
shall be achieved in a direction orthogonal to that.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
56 of 73
5.11.1. Inputs
5.11.1.1. Ellipse Fit Method
attitude_id
earth_pixel (taken from earth_pixel_list)
satellite_to_target_sc_frame (azimuth and elevation)
satellite_to_target_topo_frame (azimuth and elevation)
local north angle
W is the apodisation window term for each (u,v) baseline
Auxiliary Data
•
Default PLM values (SM_xxxx_AUX_PLM___<ID>)
f0 is the central frequency
5.11.1.2. Analytical Pixel Footprint
attitude_id
earth_pixel (taken from earth_pixel_list)
satellite_to_target_sc_frame (azimuth and elevation)
satellite_to_target_topo_frame (azimuth and elevation)
matrix_BFP2EF
satellite_position_EF
5.11.2. Outputs
Footprint_maximum
Footprint_minimum
5.11.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
Variable name
Description
Definition Type
Class Unit
Size
attitude_id
EE CFI pointing structure Section
5.2.3
xp_attit
ude_id
I
N/A
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
1
SMOS L1 Processor L1c Data
Processing Model
Variable name
Description
Definition
earth_pixel
Pixel extracted from the Section
list of pixels contained 5.5.3
inside the EAF-FOV
border
satellite_to_target_sc_f
rame
S/C to pixel directions Section
expressed in the S/C 5.6.3
frame
satellite_to_target_topo
_frame
Code
:
Date
:
29/11/2012
Issue
:
2.11
Page
:
57 of 73
Class
Unit
Size
I
N/A
1
double
I
deg
2
S/C to pixel directions Section
expressed in topocentric 5.6.3
frame
double
I
deg
2
local_north_angle
Angle
between
S/C Section 5.4
ground
track
instantaneous direction
and local north
double
I
deg
1
w
Apodisation term
double
I
N/A
1396
B
Frequency
(19MHz)
Bandwidth Section
3.3.2
double
I
MHz
1
Center_frequency
Central frequency of Section
operation
of
the 3.3.2
instrument
double
I
MHz
1
u_v
Set of non-redundant
baselines (0 – redundant,
1 – non-redundant)
t_l1c_ba I
selines
N/A
1396
matrix_BFP2EF
Rotation matrix between Section 5.2
the BFP and the EF
frame.
double
I
N/A
9
satellite_position_EF
Position of the satellite in N/A
the EF frame
double
I
N/A
3
Footprint_maximum
Maximum size of
projected elliptical
footprint
N/A
double
O
km
1
Footprint_minimum
Minimum size of
projected elliptical
footprint
N/A
double
O
km
1
Eq. 26
section
7.2.4
Type
SO-DS-DME-L1PP-0009
Table 17: Compute Pixel Footprint variable list
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
58 of 73
5.11.4. Implementation
From L1PP 1.5, there are two selectable methods that can be used to compute the footprint axes. The
first one is the original method that requires the projection of a minimum of 6 points of the footprint in
order to be fit into an ellipse. The second one computes the analytical expression of the footprint doing
certain approximations, and uses the analytical formula to derive the footprint axes.
5.11.4.1. Ellipse Fit Method
There are two main computations to be performed here. The first one shall be used to identify the xi-eta
coordinates of the two points in the –3dB contour of the Array Factor, one along the S/C to pixel
azimuth direction, and the other in the orthogonal one.
The second step shall project those points over the Earth, in order to compute the geodetic distance
between the pixel centre and them.
The first step must be done by approximating the footprint shape on the ground by an ellipse. For this
fit, 6 projected points are the minimum required.
Initially, we must use equation 10 for the computation of the Array Factor:
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7.
The six azimuth directions over which the AF needs to be computed can be spaced 60º in order to
provide maximum coverage:
φF = i × 60º (0 ≤ i ≤ 5)
Eq. 49
And the equation that needs to be solved is the following one, in order to find ∆r for each of the
previous azimuth angles:
∑∑W ( u
m
n
2
mn
, vmn )
 u ⋅ ξ + vmn ⋅η  j 2π (umn ∆r sin φ p + vmn ∆r cosφ p )
= ∑∑W ( umn , vmn ) ⋅ r%  − mn
⋅e
f0
m n


Eq. 50
 −π W 2 ( uξ + vη )2 
r



, where Wr is computed as B/f0, and
The FWF function can be approximated by r% ≈ e 
both of these values are taken from the AUX_PLM ADF. The calibrated FWF shape product could
also be used here, although the complexity increase for the gain in the accuracy computation does
not make it worthwhile.
The equation is easily solved by using the secant method, using the maximum at ∆r=0, and as start
value of ∆r half of the Blackman ideal –3dB contour. This Blackman ideal –3dB contour value is
computed as:
∆rBlackman −3dB =
2.321919
(2
3 ⋅ NEL ⋅ d
)
0.972282
Eq. 51
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
59 of 73
In the ideal case, and if the Blackman window has been used as apodisation window, the value
presented before can be used as ∆r, in order to compute the xi-eta coordinates of the six looked-for
values
ξ F = ξ + ∆rF sin φF 

ηF = η + ∆rF cos φF 
Eq. 52
Once the values have been found, the projection on Earth is done using functions already explained
before:
1. For each xi-eta point extract the corresponding azimuth and elevation angles. Azimuth must be
converted to [0,360] range before being used by CFI functions.
2. Call xp_target_inter function with attitude_id structure, the azimuth and elevation angles, deriv
XP_NO_DER and inter_flag XP_INTER_1ST, in order to obtain an initialised target_id structure.
Parameter dem_id and atmos_id shall be left NULL.
3. Call xp_target_extra_main with the target_id structure obtained in the previous step, select choice
XP_TARG_EXTRA_MAIN_GEO, target_number shall be the num_user_target returned from the
previous function and target_type XP_LOS_TARGET_TYPE, in order to obtain several output
vectors (main_results, main_results_rate, main_results_rate_rate)
4. The main_results vector contains the geodetic latitude and altitude, as well as geodetic and
geocentric longitude.
5. The geodetic latitude and the geocentric longitude shall be used in another call to xl_geod_distance
function together with the latitude and longitude of the pixel centre, in order to retrieve the geodetic
distance and the relative azimuth from the centre to each point.
Once the distance and relative azimuths for each of the 6 points is known, they can be used in any least
squares method to fit the 6 parameters of an ellipse quadratic equation, for example [RA.02].
Once the ellipse equation is known, obtaining both semi-axes is trivial.
5.11.4.2. Analytical Footprint Method
This method computes the analytical equation of the 3dB contour over the Earth’s surface, and through
that equation (approximated by an ellipse) the semi-axis and orientation can be computed.
First the equation of the aligned cone in the red reference frame of Figure 7 must be obtained. For that
purpose, two rotations are performed, first the axis’ cone is aligned with the z10y1 plane and then a
second rotation is done around the x1 axis, with an angle of (θ - 90º).
This creates a referential where the cone has the following equation:
2 2
x 22+ cos 2 (θ ) y22 − r−3dB
z 2= 0
Eq. 53
Using the rotation matrix between the BFP and the EF frame and with the CFI function
XL_EF_TO_TOPOCENTRIC this equation is transformed to the topocentric frame of the target as a
conic section. From it the major and minor semi axes of the ellipse are derived.
For further details on the implementation of this algorithm, please refer to [RD.13].
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
60 of 73
5.11.5. Error Handling
In case any error is returned by the CFI functions, the processing of this pixel shall be stopped, although
the processing for the remaining pixels in the scene can continue.
5.12. Compute Pixel Flags
This computation shall be performed for each Earth Fixed pixel belonging to the scene, in order to flag
it for different position within the FOV, or for corrections that were performed along the processing.
As a starting point, the pixel will inherit the L1b flags of the whole scene. These include the polarisation
mode, and the corrections for direct Sun, direct Moon and reflected Sun.
Afterwards, the pixel shall be flagged according to its position in the FOV (if it is inside the strict AFFOV or not, or if it is very close to the border of the Earth aliases, or to the Unit Circle aliases (also
known as the “suspenders and belt”). Information about the RFI contamination shall be taken from the
AUX_RFI ADF and from the L1b flags.
Finally, if the pixel is located very close to a Sun or Moon alias, it shall be flagged as well, and also if it
is potentially affected by Sunglint.
5.12.1. Inputs
L1b data (SM_xxxx_ MIR_SC_D1B<ID> or SM_xxxx_ MIR_SC_F1B<ID>)
•
l1b_flags
•
position of Sun and Moon in xi-eta coordinates
Auxiliary Data
•
RFI List of sources with Latitude, Longitude, BT and action to take in the flagging routines
(SM_xxxx_AUX_RFILST_<ID>).
•
Bi-static scattering coefficients (SM_xxxx_AUX_BSCAT_<ID>)
earth_pixel (taken from earth_pixel_list)
satellite_to_target_sc_frame (azimuth and elevation)
earth_eaf_fov_list (projected EAF-FOV, as computed in section 5.5)
5.12.2. Outputs
pixel_l1c_flags
5.12.3. List of variables
The following table describes the variables used in the subsequent implementation section. Variables are
listed as input, local and output (I, L, O). The Size column indicates the number of elements constituting
that variable, and NOT the size of the variable in bytes (this information can be taken from the Type
column).
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Variable name
Description
Definition
earth_pixel
Pixel extracted from the Section
list of pixels contained 5.5.3
inside the EAF-FOV
border
satellite_to_target_sc_f
rame
S/C to pixel directions Section
expressed in the S/C 5.6.3
frame
l1b_flags
Code
:
Date
:
29/11/2012
Issue
:
2.11
Page
:
61 of 73
Class Unit
Size
I
N/A
1
double
I
deg
2
Scene flags taken from N/A
the L1b scene being
processed
bit
I
N/A
8
Sun_position
Xi-eta position of the Sun N/A
in the antenna frame
double
I
N/A
2
Moon_position
Xi-eta position of the N/A
Moon in the antenna
frame
double
I
N/A
2
Earth_eaf_fov_list
List of lat-lon points in Section
the
border
of
the 5.4.3
projected EAF-FOV
I
N/A
1
rfi_pixel_list
Geographic coordinates
for RFIs listed in
AUX_RFILST
double
I
N/A
[Nbr_RF
I_in_Sna
pshot]
pixel_l1c_flags
Pixel flags computed
bit
O
N/A
16
N/A
Type
SO-DS-DME-L1PP-0009
Table 18: Compute Pixel Flags variable list
5.12.4. Implementation
The computation of the pixel_l1c_flags should begin with the simple transfer of the l1b_flags of the
scene being processed.
Another simple task is to extract from the RFI ADF the flag relative to the pixel being processed, using
the pixel unique identifier.
All pixels being processed here will always be by definition inside the EAF-FOV. In order to check id
the pixel is inside the strict AF-FOV, the following steps must be followed:
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7.
For all the image replica centres, as described in equation 28, the distance between all centres and
the xi-eta point shall be computed
If all distances are greater than 1, then the pixel is inside the strict AF-FOV. If any of the 6
computed distances is smaller than 1, then the pixel is outside the strict AF-FOV
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
62 of 73
In order to check if the pixel is close to the border, a configurable distance value must be used (it shall
be defined in the L1 algorithm configuration file by the FOV_Border_Flag_Size flag). The computation
must be performed using the pixel xi-eta coordinates, and projected EAF-FOV contour xi-eta
coordinates.
The geometric distance between the pixel and each line formed by two consecutive points in the EAFFOV contour shall be computed, using the formulation:
d = (ξ EAF − ξ pix ) 2 − (η EAF − η pix ) 2
Eq. 54
The condition that must be fulfilled to guarantee that the pixel is closer than 0.01 to the pixel
(configurable) is:
The computed distance d is smaller than 0.01
In addition, in order to check if the pixel is close to the Unit Circle Border aliases (suspenders and belt)
the following procedure will be taken:
The xi-eta coordinates for the 3rd to 5th unit circle replicas are computed, according with the
formulation:
α replica _ i =
ξ replica _ i =
ηreplica _ i =
π
6
+i
π
3
, with i =3,4,5
Eq. 54 a)
2 cos(α replica _ i )
3DisRat
with i =3,4,5
2sin(α replica _ i )
3DisRat
The distance between the replicas border and the pixel is computed:
d replica _ i = (ξ replica _ i − ξ pix ) 2 − (η replica _ i − η pix ) 2 with i =3,4,5
Eq. 54 b)
The conditions that must be fulfilled to guarantee that the pixel is closer than 0.01 of one of the
suspenders or belt are:
d replica _ i > 1 − FOV_Border_Flag_Size flag
d replica _ i < 1 + FOV_Border_Flag_Size flag
with i =3,4,5
Eq. 54 c)
Therefore the width of the suspenders and belt is twice the FOV_Border_Flag_Size flag while the
width of the border is FOV_Border_Flag_Size flag .
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
63 of 73
In order to compute the flag due to Sun or Moon alias being too close to the pixel, the following steps
must be followed:
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7.
The Sun or Moon xi.eta coordinates are extracted from the L1b scene being processed. These
coordinates represent the real xi-eta coordinates in the unit circle.
For all the image replica centres, as described in equation 28, the distance between the pixel and the
point located at replica centre plus Sun/Moon xi-eta coordinates shall be computed. This will
compute the xi-eta coordinates of the aliases that fall inside the unit circle, but come from an aliased
image.
If the distance is smaller than a given threshold (0.075/0.01 for the Sun/Moon, computed
considering a 1º beam around the Sun centre), then the pixel shall be flagged with the Sun/Moon
alias flag
The Sun alias threshold is configurable using the Sun_Point_Flag_Size of the CNFL1P configuration
file.
For the flagging of the pixels that could be affected by the Sun tails, the following image shall be taken
into consideration. This figure shows the expected propagation path of the tails due to the Hexagonal
Discrete Fourier Transform applied to L1b data:
Figure 14: Sun tails areas
The blue circle represents the aliased Sun entering in the nominal reconstruction hexagon. The green
areas represent the expected paths of ripple propagation due to a strong source in the blue circle. The
radius of the blue circle and the width of the green areas are the same as presented before (0.01 units in
xi-eta, equivalent to a 1º beam diameter). The method to compute the Sun tails flag would then be the
following:
The pixel xi-eta coordinates are extracted from the azimuth (φ) and elevation (θ) observation angles
expressed in the S/C frame, according to equation 7.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
64 of 73
The Sun xi.eta coordinates are extracted from the L1b scene being processed. These coordinates
represent the real xi-eta coordinates in the unit circle.
For all the image replica centres, only one Sun alias will enter into the nominal hexagon area. Its
coordinates (ξ SunAlias ,η SunAlias ) , shall be computed after identifying which of the aliases falls inside
the hexagon.
The distances from the pixel xi-eta coordinates to the three lines parting from the center of the Sun
alias shall be computed, using the following equations
d1 = ξ − ξ SunAlias
d2 =
d3 =
(ξ − ξ SunAlias ) + 2 (η − ηSunAlias )
Eq. 55
5
(ξ − ξ SunAlias ) − 2 (η − ηSunAlias )
5
If any of the distances is smaller than a given threshold (0.01, computed considering a 1º beam
around the Sun centre), then the pixel shall be flagged with the Sun tails flag
For what concerns the flagging of pixels if they are potentially affected by Sunglint, the algorithm to
compute this flag is described in the following steps:
In order to detect potential Sunglint contamination, the modelled Sunglint computed in L1b
processing shall be compared against a configurable threshold. This flag must be computed in L1c
due to the inherent differences in the working pixel grids between L1b (128x128 unit circle) and L1c
(4H9 ISEA), which provide a different pixel geometry
The equation that must be implemented is almost identical to Eq.34 of DPM L1b, but with the
following modification:
TBpqsun glint (ξ ,η ) =
(
r r
)
(
r r
σ 0pp n s , ni + σ 0pq n s , ni
)
Eq. 56
4π cos θ
We are not interested in the Brightness Temperature value per se, but only in the geometric
relationship between S/C–pixel–Sun, so we take out the dependency on the Sun BT and solid angle.
This way the flagging depends solely on the predicted scattering and not on the state of the Sun.
The equation above needs to take into consideration all of the points reflected in section 2.2.2.1.6 of
DPM L1b, namely in the geometric rotation to be applied, interpolation limits, etc…
Once the value above is computed, it must be compared against the configurable threshold, and if
the limit is exceeded the L1c Sunglint area flag is set for that particular measurement and pixel.
The flagging of the RFIs has two components – the one inherited from L1b/L1a and the one computed
dynamically in L1c. For that, the RFIs seen in the current snapshot are retrieved from the AUX_RFI
ADF and stored in a smaller variable, the rfi_pixel_list. Then, for each RFI in that list:
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
65 of 73
The RFI position in (ξ, η) coordinates are computed as (ξRFI, ηRFI). If the source appears in FOV,
then all pixels contained in a circle of radius given by Eq. 57 and centred on (ξRFI, ηRFI) are marked
as affected by RFI.
r = 0.39 log (T p B , RFI ) + 0.29
Eq. 57
Note that this circle is given in pixel units, which corresponds to 2/128 in (xi, eta) coordinates.
In addition, if the tails of that source are to be flagged as well, Eq.55 is used to compute the distance
between the pixel (ξ, η) and the RFI (ξRFI, ηRFI). If the distance is smaller than the diameter of the
circle previously computed, then that pixel is marked with the RFI tails flag.
Note: the RFI tails flag is the reuse of the EAFFOV flag (since all pixels were marked
with it).
If the AUX_RFILST marks the current RFI source to be flagged as corrupt, then the corrupt flag is
also set, at a snapshot level. This flag is also activated if the sum of RFI BTs in that snapshot
exceeds
the
limit
imposed
by
the
user
in
the
configuration
file
(BT_Dual/Full_RFI_Snapshot_Flag_Threshold).
The RFI pixel flag is also raised if the Brightness Temperature computed for each pixel
measurement exceeds the limit imposed by the user in the configuration file
(BT_Dual/Full_RFI_Pixel_Flag_Threshold), and in the case of Dual pol measurements it is also
raised if the Brightness Temperature is negative.
5.12.5. Error Handling
No errors should be returned on this function, unless for some strange reason two consecutive points in
the projected EAF-FOV are the same, which would cause a division by zero in the border flag
computation. However, this would indicate a previous error, not an error in the current processing.
If this situation happens, the pixel processing should be aborted, although the remaining pixels in the
scene can still be processed.
5.13. Product Generation
L1c products shall be generated in accordance to the format described in [RD.5]. In this section it is
described how to fill each substructure within each product type.
5.13.1. MPH
See section 4.1.1 of [RD.5] for the generic description. All fields in the MPH are applicable to all L1c
products and are filled according to the following rules:
Product: Filename of the physical file generated according to the rules in section 3.1
Proc_Stage_Code: Taken from the configuration mode in which the L1PP is operating
Ref_Doc: Hardcoded in the L1PP software to reflect which reference document is applicable
Header_Size: Set to 9397
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
66 of 73
Validation_Schema: Set by the Header template function of the BinXML function
Acquisition_Station: Taken from the configuration file of L1PP
Proc_Centre: Hardcoded to “L1PP”
Proc_Time: Set to the UTC time of product generation
Proc_Version: Harcoded in the L1PP software to reflect processor version
Sensing_Start: UTC time of the first snapshot inside the Measurement Data Set
Sensing_Stop: UTC time of the last snaphsot inside the Measurement Data Set
Abs_Orbit: Absolute orbit at Sensing_Start, extracted from EE CFI function xo_time_to_orbit
Phase, Cycle, Rel_Orbit: Values taken from EE CFI function xo_orbit_rel_from_abs
X_Position, Y_Position, Z_Position, X_Velocity, Y_Velocity, Z_Velocity: Values taken from EE
CFI function xo_propag
Vector_Source: Depending on the data used for initialising the propagators it may be IG
(propagation initialised with GPS data in ancillary packet), SP (propagation initialised with SOGS
Predicted data) or SR (propagation initialised with SOGS Restituted data)
Leap_UTC: Value retrieved from the EE CFI function xl_time_get_leap_second_info
Product_Confidence: Set to NOMINAL if no errors occur during processing, or ERROR if errors 25 occur.
Total_Size: Set to the binary product size plus the Header size
5.13.2. SPH
The generic format of L1c products SPH is defined in section 4.12, 4.5.1.2 and 4.5.2.2 of [RD.5].
The SPH is different between nominal L1c products and Browse L1c products. The only difference,
however, lies in the field Incidence_Angle that is set for Browse products to indicate which incidence
angle was taken to extract the swath.
All fields in the SPH are applicable to all L1c products and are filled according to the following rules:
Abs_Orbit_Start, Start_Time_ANX_T: Values extracted from EE CFI function xo_time_to_orbit
using MPH Sensing_Start as input
Abs_Orbit_Stop, Stop_Time_ANX_T: Values extracted from EE CFI function xo_time_to_orbit
using MPH Sensing_Stop as input
UTC_at_ANX: Value extracted from EE CFI function xo_orbit_to_time using Abs_Orbit_Start and
zero seconds as input
Long_at_ANX: Value extracted from EE CFI function xo_propag_extra using Abs_Orbit_Start and
zero seconds as input
Ascending_Flag: Set to A (Ascending) if Abs_Orbit_Stop > Abs_Orbit_Start, or D (Descending) if
Abs_Orbit_Stop = Abs_Orbit_Start
Mode: Adopts different values depending on the file type.
•
(APID_DUAL) MIR_SCLD1C, MIR_SCSD1C, MIR_BWLD1C, MIR_BWSD1C
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
•
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
67 of 73
(APID_FULL) MIR_SCLF1C, MIR_SCSF1C, MIR_BWLF1C, MIR_BWSF1C
Start_Lat, Start_Long: Set to boresight coordinates of first snapshot within product
Stop_Lat, Stop_Long: Set to boresight coordinates of last snapshot within product
Incidence_Angle: Set only for Browse products. Equal to 42.5º
L1C_Err_Flag: Value set to 1 if L1C_Quality > L1C_Err_Threshold. Set to 0 otherwise
L1C_Quality: Cumulative quality check set through the processing of the product.
L1C_Err_Threshold: Value taken from L1c configuration file
Apodisation_Window: Set to value according to apodisation ADF used
RSC_Flag: Flag taken from original L1B products
5.13.3. Data Set
The product format variables are filled according to the list of variables presented in the previous
chapters. The following sections detail which are the elements that shall be set as output.
5.13.3.1. Swath_Snapshot_list
Snapshot_Time, Snapshot_ID, Snapshot_OBET: Values taken from l1b_data structure
X_Position, Y_Position, Z_Position, X_Velocity, Y_Velocity, Z_Velocity, Vector_Source: Values
taken from pvt_data structure.
Q0, Q1, Q2, Q3: Values taken from aocs_data structure
TEC: Value taken from tec variable
Geomag_F, Geomag_D, Geomag_I: Values taken from geomag_f, geomag_d and geomag_I
variables
Sun_RA, Sun_DEC: Values taken from EE CFI function xl_sun and
Sun_BT: Taken from l1b_data structure
Accuracy: Taken from l1b_data structure
Radiometric_Accuracy: Taken from bt_accuracy structure computed over the boresight direction
5.13.3.2. Temp_Swath_Dual
Grid_Point_ID: Taken from earth_pixel_list structure
Counter: Number of measurements existing for the previous particular Grid Point
Flags: Taken from l1c_flags structure
BT_Value: Taken from l1c_data structure
Radiometric_Accuracy: Taken from bt_accuracy structure
Incidence_Angle, Azimuth_Angle: Taken from pixel_observation_angles structure
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
68 of 73
Faraday_Rotation_Angle, Geometric_Rotation_Angle: Taken from ionospheric_rotation_angles
structure
Snapshot_ID: Taken from l1b_data structure
Footprint_Axis1, Footprint_Axis2: Taken from footprint_size structure
5.13.3.3. Temp_Swath_Full
Same as the previous case, except that the BT_Value is replaced by two different values:
BT_Value_Real: Taken from l1c_data structure
BT_Value_Imag: Taken from l1c_data structure
5.13.3.4. Temp_Browse
Same as Temp_Swath_Dual, except that in this case Incidence_Angle does not need to be defined
The computational strategy to generate the browse data is the following:
Browse data are generated for an incidence angle of 42.5º
For any given pixel, all measurements within ±5º are used to generate the interpolated value
Only pixels which have measurements both above and below 42.5º are considered (in order to
prevent extrapolation)
Interpolation is done through a Levemberg-Marquardt LSQ fitting of all measurements inside the
±5º range
LSQ interpolation is performed for the following values:
•
BT_Value_Real
•
BT_Value_Imag
•
Radiometric_Accuract
•
Azimuth_Angle (taking care to unwrap the phase to avoid interpolation jumps between 0 and
360º)
•
Footprint_Axis1 and Footprint_Axis2
The information for the measurement flag is obtained from a bitwise OR operation on all the flags
from the measurements within the ±5º range
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
69 of 73
6. OPEN ISSUES
The computation of the pixel footprint through the analytical method described in Sections 5.10.1.2
and 5.10.4.2 can only be performed accurately for the case of the standard Blackman apodisation
window. If the user chooses to use any other apodisation window, then the footprint value must be
computed precisely, as described in Sections 5.10.1.1 and 5.10.4.1, by setting to false the Analytical
Footprint flag in the L1-configuration file (CNF file) and activate the Compute footprint size
precisely in the L1PP configuration file.
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
70 of 73
7. ANNEX 1: GLOBAL DATA TYPES, CONSTANTS AND
FUNCTIONS
The L1PP processor uses globally defined structures to transfer instrument raw and processed data
between processing units. These structures are only for memory short-term storage, for caching
purposes, the Data Provider core component structures are used.
Since, for many Processing Units, the input and output variables are global structures, it is useful to list
them here. The L1PP global types and constants are all defined in a single header file, l1pp_defs.h.
7.1. Constants
The most common numerical constants throughout the code are referenced by pre-defined constant
variable names, in order to ease control over critical numerical quantities and improve the readability of
the code itself.
7.1.1. Instrument constants
The numerical constants that define instrument and data attributes:
Constant Name
Description
Value
N_EL
Number of elements per arm
21
MIRAS_BANDWIDTH
Equivalent receiving frequency bandwidth in Hz
19
INTEGRATION_TIME
Integration time for each snapshot
1.2
C_EFF
Coefficient that accounts for the 1-bit correlation,
oversampling and hermiticity
1.81
LICEF_NUMBER
Total number of LICEF outputs from the instrument
DISTANCE_RATIO
Distance ratio between receivers
0.875
NON_RED_BASELINES
Total number of non redundant baselines
1396
72
Table 19: Instrument constants
7.1.2. Mathematical and physical constants
The mathematical and physical numerical constants necessary for calculations:
Constant Name
PI
Description
Value of π (128 bit precision)
Value
3.141592653
5897932384
Table 20: Mathematical and physical constants
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
71 of 73
7.2. Basic types
Several basic types are defined. Their purpose is to hold logically coherent information in a selfexplanatory way.
7.2.1. t_utc_time
Structure for UTC time format:
Field name
Description
Type
Unit
Size
week_number
Week number
t_uint
weeks
1
seconds
Seconds in current
week
t_uint
seconds
1
fraction_seconds
Fractions of seconds
in current second
t_uint
1/(2^32)
seconds
1
utc_flags
Flags indicating
precision of UTC time
t_utc_solutio
n
N/A
1
Table 21– t_utc_time
7.2.2. t_l1a_aocs_data
Structure for AOCS data after conversion to engineering units:
Field name
Description
Type
Unit
Size
aocs_utc
UTC time for AOCS
data
t_utc_time
N/A
1
q
Rotational angles quaternion
double
N/A
4
angular_velocity
Angular velocity,
indexed by x, y, z
double
rad/s
3
Table 22 - t_l1a_aocs
7.2.3. t_l1a_pvt_data
Structure for PVT data after conversion to engineering units:
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
72 of 73
Field name
Description
Type
Unit
Size
pvt_utc
UTC time for PVT
data
t_utc_time
N/A
1
position
Position vector,
indexed by x, y, z
double
m
3
velocity
Velocity, indexed by
x, y, z
double
m/s
3
Table 23 - t_l1a_pvt
7.2.4. t_l1c_baselines
This structure shall contain the u_v coordinates for the star domain, as they are expected from the L1b
product, and used throughout the L1c processing.
The coordinates shall be computed following the next figure, starting from the centre of the star and
moving left to right and bottom to top:
Figure 15: u_v baselines ordering
This ordering is based on reporting only the baselines with positive v coordinate and u positive for v=0:
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.
SMOS L1 Processor L1c Data
Processing Model
Code
:
SO-DS-DME-L1PP-0009
Date
:
29/11/2012
Issue
:
2.11
Page
:
73 of 73
The v coordinate for the upper half of the baselines goes continuously from 0 to sqrt(3)*NEL*d,
where NEL=21 and d=0.875, in incremental steps of sqrt(3)*d/2
The u coordinate of the upper half of the baselines shall follow the mathematical rules defined as:
•
If v=0, then u goes from d to 24*d in incremental steps of d
•
If v>0 and v<=sqrt(3)*NEL*d/2, then u goes from –(NEL*d +v/sqrt(3)) to +(NEL*d
+v/sqrt(3)) in incremental steps of d
•
If v=sqrt(3)*(NEL+1)*d/2, then u goes from –11*d to +11*d in incremental steps of d
•
If v=sqrt(3)*(NEL+2)*d/2, then u has the values –23*d/2, –19*d/2 to +19*d/2 in incremental
steps of d and +23*d/2. Notice that the elements ±21*d/2 are not present.
•
If v=sqrt(3)*(NEL+3)*d/2, then u has the values –12*d, –9*d to +9*d in incremental steps of
d and +12*d. Notice that the elements ±11*d and ±10*d are not present.
•
Finally, if v>sqrt(3)*(NEL+3)*d/2 and v<=sqrt(3)*NEL*d, then u goes from –(NEL*d –
v/sqrt(3)) to +(NEL*d –v/sqrt(3)) in incremental steps of d
This document is property of DEIMOS Engenharia and cannot be distributed or duplicated without its written permission.